From 80059eaa3b97670715eec294b9342b4ff8abe4ac Mon Sep 17 00:00:00 2001
From: jefyang1 <146495389+jefyang1@users.noreply.github.com>
Date: Tue, 4 Mar 2025 10:32:25 -0800
Subject: [PATCH] Remove CK_USE_AMD_MFMA_GFX950 (#1935)

* Add runtime check in example_gemm_xdl_streamk for gfx950

* Add runtime check in grouped conv fwd examples for gfx950

* Disable CK_USE_AMD_MFMA_GFX950

* Add new instances for gfx950

* Fix test_gemm_universal on gfx950

[ROCm/composable_kernel commit: c95bda93bacd4b8f809becd7c4b9e8b743d0ded8]
---
 CMakeLists.txt                                |   3 -
 example/01_gemm/gemm_xdl_streamk.cpp          |  18 +-
 example/01_gemm/run_gemm_example.inc          |  78 +----
 example/01_gemm/run_gemm_example_streamk.inc  | 270 ++++++++++++++++++
 ...grouped_conv_fwd_bias_relu_add_example.inc | 100 ++++---
 ...iple_d_welford_first_half_xdl_cshuffle.hpp |   5 +-
 ...wise_batched_gemm_gemm_xdl_cshuffle_v1.hpp |   5 +-
 ...iple_d_gemm_multiple_d_xdl_cshuffle_v1.hpp |  11 +-
 ...ultiple_d_softmax_gemm_xdl_cshuffle_v1.hpp |  11 +-
 ...ched_gemm_softmax_gemm_xdl_cshuffle_v1.hpp |   5 +-
 ...e_gemm_bias_add_reduce_xdl_cshuffle_v1.hpp |   5 +-
 ...ridwise_gemm_multiple_abd_xdl_cshuffle.hpp |  20 +-
 ...emm_multiple_d_multiple_r_xdl_cshuffle.hpp |   5 +-
 .../gridwise_gemm_multiple_d_xdl_cshuffle.hpp |  20 +-
 ...ultiple_d_xdl_cshuffle_lds_direct_load.hpp |  20 +-
 ...se_gemm_multiple_d_xdl_splitk_cshuffle.hpp |   5 +-
 .../gridwise_gemm_reduce_xdl_cshuffle_v1.hpp  |   5 +-
 ...e_gemm_split_k_multiple_d_xdl_cshuffle.hpp |  10 +-
 ...emm_split_k_multiple_d_xdl_cshuffle_v2.hpp |   5 +-
 ...idwise_gemm_xdl_cshuffle_bwd_weight_v3.hpp |   5 +-
 .../gridwise_gemm_xdl_cshuffle_streamk_v3.hpp |   5 +-
 .../grid/gridwise_gemm_xdl_cshuffle_v1.hpp    |  12 +-
 .../grid/gridwise_gemm_xdl_cshuffle_v2.hpp    |   5 +-
 .../grid/gridwise_gemm_xdl_cshuffle_v3.hpp    |   5 +-
 .../gridwise_gemm_xdl_cshuffle_v3_b_scale.hpp |   5 +-
 ...ridwise_gemm_xdl_cshuffle_v3_multi_abd.hpp |  14 +-
 .../gridwise_gemm_xdl_cshuffle_v3_multi_d.hpp |  15 +-
 ..._gemm_xdl_cshuffle_v3_multi_d_ab_scale.hpp |  15 +-
 ...ridwise_gemm_xdl_layernorm_cshuffle_v1.hpp |   5 +-
 ...ridwise_gemm_xdl_waveletmodel_cshuffle.hpp |   5 +-
 .../grid/gridwise_gemm_xdlops_bwd_weight.hpp  |  14 +-
 .../gpu/grid/gridwise_gemm_xdlops_v3r1.hpp    |   5 +-
 .../tensor_operation/gpu/warp/xdlops_gemm.hpp |  76 ++---
 ...conv_bwd_weight_two_stage_xdl_instance.hpp |  32 +--
 ...ice_grouped_conv_fwd_xdl_comp_instance.hpp | 117 ++++++--
 ...ed_conv_fwd_xdl_merged_groups_instance.hpp |  56 +++-
 ...m_xdl_f16_f16_f16_gkm_gkn_gmn_instance.cpp |  23 +-
 ...m_xdl_f16_f16_f16_gkm_gnk_gmn_instance.cpp |  23 +-
 ...m_xdl_f16_f16_f16_gmk_gkn_gmn_instance.cpp |  40 ++-
 ...m_xdl_f16_f16_f16_gmk_gnk_gmn_instance.cpp |  40 ++-
 ...6_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp |   8 +-
 ...f16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp |   4 +-
 ...6_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp |   4 +-
 ...f16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp |   4 +-
 ...6_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp |  46 ++-
 ..._c_shuffle_nhwc_kyxc_nhwk_f16_instance.cpp |  89 ++++--
 ..._bias_relu_nhwc_kyxc_nhwk_f16_instance.cpp |  95 ++++--
 ...s_relu_add_nhwc_kyxc_nhwk_f16_instance.cpp |  96 +++++--
 ..._2_stage_f16_f16_f16_mk_nk_mn_instance.cpp |  23 +-
 ...uffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp |  32 ++-
 ..._shuffle_f16_f16_f16_km_kn_mn_instance.cpp |  23 +-
 ..._shuffle_f16_f16_f16_km_nk_mn_instance.cpp |  23 +-
 ..._shuffle_f16_f16_f16_mk_kn_mn_instance.cpp |  25 +-
 ..._shuffle_f16_f16_f16_mk_nk_mn_instance.cpp |  28 +-
 ...l_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp |  23 +-
 ...m_kn_mn_interwave_pipeline_v1_instance.cpp |   9 +-
 ...regular_interwave_pipeline_v1_instance.cpp |   2 -
 ...m_nk_mn_interwave_pipeline_v1_instance.cpp |   9 +-
 ...regular_interwave_pipeline_v1_instance.cpp |   2 -
 ...k_kn_mn_interwave_pipeline_v1_instance.cpp |   9 +-
 ...regular_interwave_pipeline_v1_instance.cpp |   2 -
 ...k_nk_mn_interwave_pipeline_v1_instance.cpp |   2 -
 ...regular_interwave_pipeline_v1_instance.cpp |   2 -
 ...16_f16_f16_f16_km_kn_mn_mn_mn_instance.cpp |  74 +++--
 ...16_f16_f16_f16_km_nk_mn_mn_mn_instance.cpp |  74 +++--
 ...16_f16_f16_f16_mk_kn_mn_mn_mn_instance.cpp |  74 +++--
 ...16_f16_f16_f16_mk_nk_mn_mn_mn_instance.cpp |  74 +++--
 ...e_f16_f16_f16_f16_km_kn_mn_mn_instance.cpp |  65 ++++-
 ...e_f16_f16_f16_f16_km_nk_mn_mn_instance.cpp |  65 ++++-
 ...e_f16_f16_f16_f16_mk_kn_mn_mn_instance.cpp |  65 ++++-
 ...e_f16_f16_f16_f16_mk_nk_mn_mn_instance.cpp |  65 ++++-
 ..._layernorm_f16_km_kn_mn_mn_mn_instance.cpp |   8 +-
 ..._layernorm_f16_km_nk_mn_mn_mn_instance.cpp |   8 +-
 ..._layernorm_f16_mk_kn_mn_mn_mn_instance.cpp |   8 +-
 ..._layernorm_f16_mk_nk_mn_mn_mn_instance.cpp |   8 +-
 ..._shuffle_f16_f16_f16_km_kn_mn_instance.cpp |  71 +++--
 ..._shuffle_f16_f16_f16_km_nk_mn_instance.cpp |  71 +++--
 ..._shuffle_f16_f16_f16_mk_kn_mn_instance.cpp |  63 +++-
 ..._shuffle_f16_f16_f16_mk_nk_mn_instance.cpp |  63 +++-
 ...f16_f16_mk_kn_mn_v1_interwave_instance.cpp |   4 +-
 ...f16_f16_mk_kn_mn_v1_irregular_instance.cpp |   4 +-
 ...f16_f16_mk_nk_mn_v1_interwave_instance.cpp |   4 +-
 ..._xdl_universal_bf16_bf16_bf16_km_kn_mn.hpp |   8 +-
 ..._xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp |  32 ++-
 ...16_bf16_km_nk_mn_comp_default_instance.cpp |   8 +
 ...6_bf16_km_nk_mn_comp_kpadding_instance.cpp |   8 +
 ..._bf16_km_nk_mn_comp_mkpadding_instance.cpp |   9 +
 ...6_bf16_km_nk_mn_comp_mpadding_instance.cpp |   8 +
 ..._bf16_km_nk_mn_mem_v1_default_instance.cpp |   9 +
 ...bf16_km_nk_mn_mem_v1_kpadding_instance.cpp |   9 +
 ...f16_km_nk_mn_mem_v1_mkpadding_instance.cpp |   9 +
 ..._bf16_km_nk_mn_mem_v2_default_instance.cpp |   9 +
 ...bf16_km_nk_mn_mem_v2_kpadding_instance.cpp |   9 +
 ...f16_km_nk_mn_mem_v2_mkpadding_instance.cpp |   9 +
 ..._xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp |  35 +--
 ...16_bf16_mk_kn_mn_comp_default_instance.cpp |   8 +
 ...6_bf16_mk_kn_mn_comp_kpadding_instance.cpp |   8 +
 ...bf16_mk_kn_mn_comp_mnkpadding_instance.cpp |   9 +
 ..._bf16_mk_kn_mn_comp_mnpadding_instance.cpp |   9 +
 ..._bf16_mk_kn_mn_mem_v1_default_instance.cpp |   9 +
 ...bf16_mk_kn_mn_mem_v1_kpadding_instance.cpp |   9 +
 ...16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp |  10 +
 ..._bf16_mk_kn_mn_mem_v2_default_instance.cpp |   9 +
 ...bf16_mk_kn_mn_mem_v2_kpadding_instance.cpp |   9 +
 ...16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp |  10 +
 ..._xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp |  46 +--
 ...16_bf16_mk_nk_mn_comp_default_instance.cpp |   8 +
 ...6_bf16_mk_nk_mn_comp_kpadding_instance.cpp |   8 +
 ..._bf16_mk_nk_mn_mem_v1_default_instance.cpp |   9 +
 ...bf16_mk_nk_mn_mem_v1_kpadding_instance.cpp |   9 +
 ..._bf16_mk_nk_mn_mem_v2_default_instance.cpp |   9 +
 ...bf16_mk_nk_mn_mem_v2_kpadding_instance.cpp |   9 +
 ...emm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp |  18 +-
 ...f16_f16_mk_kn_mn_comp_default_instance.cpp |   8 +
 ...16_f16_mk_kn_mn_comp_kpadding_instance.cpp |   8 +
 ..._f16_mk_kn_mn_comp_mnkpadding_instance.cpp |   8 +
 ...6_f16_mk_kn_mn_comp_mnpadding_instance.cpp |   8 +
 ...emm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp |  32 ++-
 ...f16_f16_mk_nk_mn_comp_default_instance.cpp |   8 +
 ...16_f16_mk_nk_mn_comp_kpadding_instance.cpp |   8 +
 ...6_f16_mk_nk_mn_mem_v1_default_instance.cpp |   9 +
 ..._f16_mk_nk_mn_mem_v1_kpadding_instance.cpp |   9 +
 ...6_f16_mk_nk_mn_mem_v2_default_instance.cpp |   9 +
 ..._f16_mk_nk_mn_mem_v2_kpadding_instance.cpp |   9 +
 ...gemm_xdl_universal_f16_f8_f16_mk_kn_mn.hpp |  18 +-
 ..._f8_f16_mk_kn_mn_comp_default_instance.cpp |   8 +
 ...f8_f16_mk_kn_mn_comp_kpadding_instance.cpp |   8 +
 ..._f16_mk_kn_mn_comp_mnkpadding_instance.cpp |   8 +
 ...8_f16_mk_kn_mn_comp_mnpadding_instance.cpp |   8 +
 ...gemm_xdl_universal_f16_f8_f16_mk_nk_mn.hpp |  18 +-
 ..._f8_f16_mk_nk_mn_comp_default_instance.cpp |   8 +
 ...f8_f16_mk_nk_mn_comp_kpadding_instance.cpp |   8 +
 ...gemm_xdl_universal_f8_f8_bf16_mk_kn_mn.hpp |  22 +-
 ...f8_bf16_mk_kn_mn_comp_default_instance.cpp |   8 +
 ...8_bf16_mk_kn_mn_comp_kpadding_instance.cpp |   8 +
 ..._bf16_mk_kn_mn_comp_nkpadding_instance.cpp |   8 +
 ...gemm_xdl_universal_f8_f8_bf16_mk_nk_mn.hpp |  29 +-
 ...f8_bf16_mk_nk_mn_comp_default_instance.cpp |   8 +
 ...8_bf16_mk_nk_mn_comp_kpadding_instance.cpp |   8 +
 ..._xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp |  12 +-
 ...16_bf16_mk_nk_mn_comp_default_instance.cpp |   9 +
 ..._xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp |  20 +-
 ...16_bf16_mk_kn_mn_comp_default_instance.cpp |   9 +
 ...6_bf16_mk_kn_mn_comp_kpadding_instance.cpp |   9 +
 ...bf16_mk_kn_mn_comp_mnkpadding_instance.cpp |   9 +
 ..._bf16_mk_kn_mn_comp_mnpadding_instance.cpp |   9 +
 ...mm_xdl_universal_bf16_i8_bf16_mk_kn_mn.hpp |  25 +-
 ...i8_bf16_mk_kn_mn_comp_default_instance.cpp |   9 +
 ...8_bf16_mk_kn_mn_comp_kpadding_instance.cpp |   9 +
 ...bf16_mk_kn_mn_comp_mnkpadding_instance.cpp |   9 +
 ..._bf16_mk_kn_mn_comp_mnpadding_instance.cpp |   9 +
 ...emm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp |  20 +-
 ...f16_f16_mk_kn_mn_comp_default_instance.cpp |   9 +
 ...16_f16_mk_kn_mn_comp_kpadding_instance.cpp |   9 +
 ..._f16_mk_kn_mn_comp_mnkpadding_instance.cpp |   9 +
 ...6_f16_mk_kn_mn_comp_mnpadding_instance.cpp |   9 +
 ...versal_streamk_bf16_bf16_bf16_mk_kn_mn.hpp |  10 +-
 ...16_bf16_mk_kn_mn_comp_default_instance.cpp |   9 +
 ...6_bf16_mk_kn_mn_comp_kpadding_instance.cpp |   9 +
 ...bf16_mk_kn_mn_comp_mnkpadding_instance.cpp |   9 +
 ..._bf16_mk_kn_mn_comp_mnpadding_instance.cpp |   9 +
 ...versal_streamk_bf16_bf16_bf16_mk_nk_mn.hpp |  13 +-
 ...16_bf16_mk_nk_mn_comp_default_instance.cpp |   9 +
 ...6_bf16_mk_nk_mn_comp_kpadding_instance.cpp |   9 +
 ...universal_streamk_f16_f16_f16_mk_kn_mn.hpp |  11 +-
 ...universal_streamk_f16_f16_f16_mk_nk_mn.hpp |  18 +-
 ...f16_f16_mk_nk_mn_comp_default_instance.cpp |   9 +
 ...16_f16_mk_nk_mn_comp_kpadding_instance.cpp |   9 +
 ..._f16_mk_nk_mn_comp_mnkpadding_instance.cpp |   9 +
 ...6_f16_mk_nk_mn_comp_mnpadding_instance.cpp |   9 +
 ...l_ngchw_gkyxc_ngkhw_bf16_comp_instance.cpp |  25 ++
 ...dl_ngchw_gkyxc_ngkhw_f16_comp_instance.cpp |  25 ++
 ...l_ngchw_gkyxc_ngkhw_int8_comp_instance.cpp |  25 ++
 ...l_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp |  79 +++++
 ...dl_nhwgc_gkyxc_nhwgk_f16_comp_instance.cpp |  79 +++++
 ...l_nhwgc_gkyxc_nhwgk_int8_comp_instance.cpp |  79 +++++
 ...groups_nhwgc_gkyxc_nhwgk_bf16_instance.cpp |  56 ++--
 ..._groups_nhwgc_gkyxc_nhwgk_f16_instance.cpp |  56 ++--
 ...dhwgc_gkzyxc_ndhwgk_bf16_comp_instance.cpp |  57 ++++
 ...ndhwgc_gkzyxc_ndhwgk_f16_comp_instance.cpp |  57 ++++
 ...f16_f8_f16_mk_kn_mn_irregular_instance.cpp |   4 +-
 ...le_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp |  24 +-
 ...i8_bf16_mk_kn_mn_comp_default_instance.cpp |  12 +
 ...8_bf16_mk_kn_mn_comp_kpadding_instance.cpp |  12 +
 ...bf16_mk_kn_mn_comp_mnkpadding_instance.cpp |  12 +
 ..._bf16_mk_kn_mn_comp_mnpadding_instance.cpp |  12 +
 186 files changed, 3272 insertions(+), 883 deletions(-)
 mode change 100755 => 100644 example/01_gemm/gemm_xdl_streamk.cpp
 create mode 100644 example/01_gemm/run_gemm_example_streamk.inc
 mode change 100755 => 100644 library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp
 mode change 100755 => 100644 library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp
 mode change 100755 => 100644 library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
 mode change 100755 => 100644 library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
 mode change 100755 => 100644 library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp
 mode change 100755 => 100644 library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instance.cpp
 mode change 100755 => 100644 library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn.hpp
 mode change 100755 => 100644 library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp

diff --git a/CMakeLists.txt b/CMakeLists.txt
index e90f893de0..3be508382a 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -202,9 +202,6 @@ if (SUPPORTED_GPU_TARGETS MATCHES "gfx94" OR SUPPORTED_GPU_TARGETS MATCHES "gfx9
     add_definitions(-DCK_USE_GFX94)
     set(CK_USE_GFX94 "ON")
 endif()
-if (SUPPORTED_GPU_TARGETS MATCHES "gfx95")
-	add_definitions(-DCK_USE_AMD_MFMA_GFX950)
-endif()
 if (SUPPORTED_GPU_TARGETS MATCHES "gfx11" OR SUPPORTED_GPU_TARGETS MATCHES "gfx12")
     message("Enabling WMMA instances")
     add_definitions(-DCK_USE_WMMA)
diff --git a/example/01_gemm/gemm_xdl_streamk.cpp b/example/01_gemm/gemm_xdl_streamk.cpp
old mode 100755
new mode 100644
index 01542c4775..41665a79b7
--- a/example/01_gemm/gemm_xdl_streamk.cpp
+++ b/example/01_gemm/gemm_xdl_streamk.cpp
@@ -27,22 +27,24 @@ using DeviceGemmStreamK = ck::tensor_operation::device::DeviceGemmXdlStreamK
 // ######|      Type|      Type|      Type|        Type|        |        |        | Elementwise| Elementwise| Elementwise|   Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
 // ######|          |          |          |            |        |        |        |   Operation|   Operation|   Operation|       |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
 // ######|          |          |          |            |        |        |        |            |            |            |       |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-    < ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout,  AElementOp,  BElementOp,  CElementOp,    256,   256,   128,     4, 16,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,      S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8>;
-#else // defined(CK_USE_AMD_MFMA_GFX950)
     < ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout,  AElementOp,  BElementOp,  CElementOp,    256,   128,   128,     4,  8,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,      S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8>;
 
           // < ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout,  AElementOp,  BElementOp,  CElementOp,    256,   256,   128,     4,   8,  32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              2,         1,           1,           1,               S<1, 32, 1, 8>,              8>;
       //  < ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout,  AElementOp,  BElementOp,  CElementOp,    128,   32,   64,     4,   8,  32,   32,    1,    1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 16, 1, 8>,              8>;
        // < ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout,  AElementOp,  BElementOp,  CElementOp,    128,   32,   128,     4,   8,  32,   32,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              2,         1,           1,           1,               S<1, 32, 1, 4>,              8>;
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
 
+// instance for double rate mfma instruction on gfx950
+using DeviceGemmStreamK2 = ck::tensor_operation::device::DeviceGemmXdlStreamK
+// ######|     AData|     BData|     CData|     AccData| ALayout| BLayout| CLayout|           A|           B|           C|  Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+// ######|      Type|      Type|      Type|        Type|        |        |        | Elementwise| Elementwise| Elementwise|   Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+// ######|          |          |          |            |        |        |        |   Operation|   Operation|   Operation|       |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+// ######|          |          |          |            |        |        |        |            |            |            |       |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+    < ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout,  AElementOp,  BElementOp,  CElementOp,    256,   256,   128,     4, 16,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,      S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8>;
 
-
-// // clang-format on
 // clang-format on
 
-using DeviceGemmInstance = DeviceGemmStreamK;
+using DeviceGemmInstance  = DeviceGemmStreamK;
+using DeviceGemmInstance2 = DeviceGemmStreamK2;
 
 using ReferenceGemmInstance = ck::tensor_operation::host::
     ReferenceGemm<ADataType, BDataType, CDataType, AccDataType, AElementOp, BElementOp, CElementOp>;
@@ -58,6 +60,6 @@ using ReferenceGemmInstanceGPU = ck::tensor_operation::device::ReferenceGemm<ALa
                                                                              BElementOp,
                                                                              CElementOp>;
 
-#include "run_gemm_example.inc"
+#include "run_gemm_example_streamk.inc"
 
 int main(int argc, char* argv[]) { return !run_gemm_streamk_example(argc, argv); }
diff --git a/example/01_gemm/run_gemm_example.inc b/example/01_gemm/run_gemm_example.inc
index 4371af6244..c064ed500c 100644
--- a/example/01_gemm/run_gemm_example.inc
+++ b/example/01_gemm/run_gemm_example.inc
@@ -3,8 +3,6 @@
 
 #pragma once
 
-#include "ck/tensor_operation/gpu/device/device_gemm_streamk.hpp"
-
 template <typename ProblemType>
 bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
 {
@@ -124,23 +122,12 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
     auto b_element_op = BElementOp{};
     auto c_element_op = CElementOp{};
 
-    using BaseStreamK = ck::tensor_operation::device::DeviceGemmStreamK<ALayout,
-                                                                        BLayout,
-                                                                        CLayout,
-                                                                        ADataType,
-                                                                        BDataType,
-                                                                        CDataType,
-                                                                        AElementOp,
-                                                                        BElementOp,
-                                                                        CElementOp>;
-
     // do GEMM
     auto gemm      = DeviceGemmInstance{};
     auto invoker   = gemm.MakeInvoker();
     float ave_time = 0;
 
-    if constexpr(std::is_same<ProblemType, ProblemSize>::value &&
-                 !std::is_base_of<BaseStreamK, DeviceGemmInstance>::value)
+    if constexpr(std::is_same<ProblemType, ProblemSize>::value)
     {
         auto argument = gemm.MakeArgument(
 #ifdef BUILD_INT4_EXAMPLE
@@ -171,61 +158,6 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
 
         ave_time = invoker.Run(argument, StreamConfig{nullptr, config.time_kernel});
     }
-    else if constexpr(std::is_same<ProblemType, ProblemSizeStreamK>::value &&
-                      std::is_base_of<BaseStreamK, DeviceGemmInstance>::value)
-    {
-        auto argument = gemm.MakeArgument(
-#ifdef BUILD_INT4_EXAMPLE
-            static_cast<KernelADataType*>(a_m_k_device_buf.GetDeviceBuffer()),
-            static_cast<KernelBDataType*>(b_k_n_device_buf.GetDeviceBuffer()),
-            static_cast<KernelCDataType*>(c_m_n_device_buf.GetDeviceBuffer()),
-#else
-            static_cast<ADataType*>(a_m_k_device_buf.GetDeviceBuffer()),
-            static_cast<BDataType*>(b_k_n_device_buf.GetDeviceBuffer()),
-            static_cast<CDataType*>(c_m_n_device_buf.GetDeviceBuffer()),
-#endif
-            M,
-            N,
-            K,
-            StrideA,
-            StrideB,
-            StrideC,
-            a_element_op,
-            b_element_op,
-            c_element_op,
-            problem_size.NumSKBlocks);
-
-        if(!gemm.IsSupportedArgument(argument))
-        {
-            std::cerr << gemm.GetTypeString() << " does not support this problem" << std::endl;
-
-            return true;
-        }
-
-        std::size_t workspace_size = gemm.GetWorkSpaceSize(&argument);
-        if(workspace_size != 0)
-        {
-            workspace.Realloc(workspace_size);
-            gemm.SetWorkSpacePointer(&argument, workspace.GetDeviceBuffer());
-        }
-
-        ave_time = invoker.Run(argument, StreamConfig{nullptr, config.time_kernel});
-
-#if 0
-        // TODO!!!!!
-        if(workspace_size != 0){
-            float * ws_ptr = reinterpret_cast<float*>(malloc(workspace_size));
-            size_t ws_dwords = workspace_size / sizeof(float);
-            workspace.FromDevice(ws_ptr);
-
-            for(size_t i = 0; i < ws_dwords; i++) {
-                uint32_t rere = reinterpret_cast<uint32_t*>(ws_ptr)[i];
-                printf("%4lu : %f(0x%08x)\n", i, ws_ptr[i], rere);
-            }
-            free(ws_ptr);
-        }
-#endif
-    }
     else
     {
         // When the Problem Type and Problem Size does not fit.
@@ -319,11 +251,3 @@ bool run_gemm_example(int argc, char* argv[])
 
     return !parse_cmd_args(argc, argv, problem_size, config) || run_gemm(problem_size, config);
 }
-
-bool run_gemm_streamk_example(int argc, char* argv[])
-{
-    ProblemSizeStreamK problem_size;
-    ExecutionConfig config;
-
-    return !parse_cmd_args(argc, argv, problem_size, config) || run_gemm(problem_size, config);
-}
diff --git a/example/01_gemm/run_gemm_example_streamk.inc b/example/01_gemm/run_gemm_example_streamk.inc
new file mode 100644
index 0000000000..438afcf71a
--- /dev/null
+++ b/example/01_gemm/run_gemm_example_streamk.inc
@@ -0,0 +1,270 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include "ck/host_utility/device_prop.hpp"
+#include "ck/tensor_operation/gpu/device/device_gemm_streamk.hpp"
+
+template <typename ProblemType>
+bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
+{
+#if defined(BUILD_INT4_EXAMPLE) && defined(CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4)
+    static_assert(sizeof(ck::int4_t) == sizeof(int8_t));
+#endif
+
+    using namespace ck::literals;
+
+    auto M       = problem_size.M;
+    auto N       = problem_size.N;
+    auto K       = problem_size.K;
+    auto StrideA = problem_size.StrideA;
+    auto StrideB = problem_size.StrideB;
+    auto StrideC = problem_size.StrideC;
+
+    auto f_host_tensor_descriptor =
+        [](std::size_t row, std::size_t col, std::size_t stride, auto layout) {
+            if constexpr(std::is_same_v<decltype(layout), ck::tensor_layout::gemm::RowMajor>)
+            {
+                return HostTensorDescriptor({row, col}, {stride, 1_uz});
+            }
+            else
+            {
+                return HostTensorDescriptor({row, col}, {1_uz, stride});
+            }
+        };
+
+    auto f_get_default_stride =
+        [](std::size_t row, std::size_t col, ck::index_t stride, auto layout) {
+            if(stride == -1)
+            {
+                // give a chance if stride is -1, return a default packed stride
+                if constexpr(std::is_same_v<decltype(layout), ck::tensor_layout::gemm::RowMajor>)
+                {
+                    return static_cast<std::size_t>(col);
+                }
+                else
+                {
+                    return static_cast<std::size_t>(row);
+                }
+            }
+            else
+                return static_cast<std::size_t>(stride);
+        };
+
+    StrideA = f_get_default_stride(M, K, StrideA, ALayout{});
+    StrideB = f_get_default_stride(K, N, StrideB, BLayout{});
+    StrideC = f_get_default_stride(M, N, StrideC, CLayout{});
+
+    Tensor<ADataType> a_m_k(f_host_tensor_descriptor(M, K, StrideA, ALayout{}));
+    Tensor<BDataType> b_k_n(f_host_tensor_descriptor(K, N, StrideB, BLayout{}));
+
+    switch(config.init_method)
+    {
+    case 0:
+        ck::utils::FillConstant<ADataType>{ck::type_convert<ADataType>(1.f)}(a_m_k);
+        ck::utils::FillConstant<BDataType>{ck::type_convert<BDataType>(1.f)}(b_k_n);
+        break;
+    case 1:
+        ck::utils::FillUniformDistributionIntegerValue<ADataType>{-5.f, 5.f}(a_m_k);
+        ck::utils::FillUniformDistributionIntegerValue<BDataType>{-5.f, 5.f}(b_k_n);
+        break;
+    case 2:
+        ck::utils::FillUniformDistribution<ADataType>{-1.f, 1.f}(a_m_k);
+        ck::utils::FillUniformDistribution<BDataType>{-1.f, 1.f}(b_k_n);
+        break;
+    case 3:
+        ck::utils::FillUniformDistributionIntegerValue<ADataType>{1.f, 1.f}(a_m_k);
+        ck::utils::FillUniformDistributionIntegerValue<BDataType>{-5.f, 5.f}(b_k_n);
+        break;
+    case 4:
+        ck::utils::FillUniformDistributionIntegerValue<ADataType>{-5.f, 5.f}(a_m_k);
+        ck::utils::FillUniformDistributionIntegerValue<BDataType>{1.f, 1.f}(b_k_n);
+        break;
+    case 5:
+        ck::utils::FillUniformDistributionIntegerValue<ADataType>{-2.f, 2.f}(a_m_k);
+        ck::utils::FillUniformDistributionIntegerValue<BDataType>{-2.f, 2.f}(b_k_n);
+        break;
+    default:
+        ck::utils::FillUniformDistribution<ADataType>{-0.1f, 0.1f}(a_m_k);
+        ck::utils::FillUniformDistribution<BDataType>{-0.1f, 0.1f}(b_k_n);
+    }
+
+    Tensor<CDataType> c_m_n_host_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
+    Tensor<CDataType> c_m_n_device_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
+    Tensor<CDataType> c_m_n_device_ref_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
+
+    std::cout << "a_m_k: " << a_m_k.mDesc << std::endl;
+    std::cout << "b_k_n: " << b_k_n.mDesc << std::endl;
+    std::cout << "c_m_n: " << c_m_n_host_result.mDesc << std::endl;
+
+#ifdef BUILD_INT4_EXAMPLE
+    DeviceMem a_m_k_device_buf(sizeof(KernelADataType) * a_m_k.mDesc.GetElementSpaceSize());
+    DeviceMem b_k_n_device_buf(sizeof(KernelBDataType) * b_k_n.mDesc.GetElementSpaceSize());
+    DeviceMem c_m_n_device_buf(sizeof(KernelCDataType) *
+                               c_m_n_device_result.mDesc.GetElementSpaceSize());
+
+    const Tensor<KernelADataType> a_m_k_converted(a_m_k);
+    const Tensor<KernelBDataType> b_k_n_converted(b_k_n);
+
+    a_m_k_device_buf.ToDevice(a_m_k_converted.mData.data());
+    b_k_n_device_buf.ToDevice(b_k_n_converted.mData.data());
+#else
+    DeviceMem a_m_k_device_buf(sizeof(ADataType) * a_m_k.mDesc.GetElementSpaceSize());
+    DeviceMem b_k_n_device_buf(sizeof(BDataType) * b_k_n.mDesc.GetElementSpaceSize());
+    DeviceMem c_m_n_device_buf(sizeof(CDataType) * c_m_n_device_result.mDesc.GetElementSpaceSize());
+    DeviceMem c_m_n_device_ref_buf(sizeof(CDataType) *
+                                   c_m_n_device_ref_result.mDesc.GetElementSpaceSize());
+
+    a_m_k_device_buf.ToDevice(a_m_k.mData.data());
+    b_k_n_device_buf.ToDevice(b_k_n.mData.data());
+#endif
+    DeviceMem workspace;
+
+    auto a_element_op = AElementOp{};
+    auto b_element_op = BElementOp{};
+    auto c_element_op = CElementOp{};
+
+    using BaseStreamK = ck::tensor_operation::device::DeviceGemmStreamK<ALayout,
+                                                                        BLayout,
+                                                                        CLayout,
+                                                                        ADataType,
+                                                                        BDataType,
+                                                                        CDataType,
+                                                                        AElementOp,
+                                                                        BElementOp,
+                                                                        CElementOp>;
+
+    // do GEMM
+    static_assert(std::is_base_of<BaseStreamK, DeviceGemmInstance>::value &&
+                  std::is_base_of<BaseStreamK, DeviceGemmInstance2>::value);
+    auto gemm           = DeviceGemmInstance{};
+    auto gemm2          = DeviceGemmInstance2{}; // instance for double rate mfma instruction
+    BaseStreamK* op_ptr = (ck::get_device_name() == "gfx950") ? static_cast<BaseStreamK*>(&gemm2)
+                                                              : static_cast<BaseStreamK*>(&gemm);
+
+    float ave_time   = 0;
+    auto invoker_ptr = op_ptr->MakeInvokerPointer();
+
+    auto argument_ptr = op_ptr->MakeArgumentPointer(
+#ifdef BUILD_INT4_EXAMPLE
+        static_cast<KernelADataType*>(a_m_k_device_buf.GetDeviceBuffer()),
+        static_cast<KernelBDataType*>(b_k_n_device_buf.GetDeviceBuffer()),
+        static_cast<KernelCDataType*>(c_m_n_device_buf.GetDeviceBuffer()),
+#else
+        static_cast<ADataType*>(a_m_k_device_buf.GetDeviceBuffer()),
+        static_cast<BDataType*>(b_k_n_device_buf.GetDeviceBuffer()),
+        static_cast<CDataType*>(c_m_n_device_buf.GetDeviceBuffer()),
+#endif
+        M,
+        N,
+        K,
+        StrideA,
+        StrideB,
+        StrideC,
+        a_element_op,
+        b_element_op,
+        c_element_op,
+        problem_size.NumSKBlocks);
+
+    if(!op_ptr->IsSupportedArgument(argument_ptr.get()))
+    {
+        std::cerr << op_ptr->GetTypeString() << " does not support this problem" << std::endl;
+
+        return true;
+    }
+
+    auto argument              = argument_ptr.get();
+    std::size_t workspace_size = op_ptr->GetWorkSpaceSize(argument);
+    if(workspace_size != 0)
+    {
+        workspace.Realloc(workspace_size);
+        op_ptr->SetWorkSpacePointer(argument, workspace.GetDeviceBuffer());
+    }
+
+    ave_time = invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, config.time_kernel});
+
+    std::size_t flop = 2_uz * M * N * K;
+    std::size_t num_btype =
+        sizeof(ADataType) * M * K + sizeof(BDataType) * K * N + sizeof(CDataType) * M * N;
+
+    float tflops = static_cast<float>(flop) / 1.E9 / ave_time;
+
+    float gb_per_sec = num_btype / 1.E6 / ave_time;
+
+    std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
+              << op_ptr->GetTypeString() << std::endl;
+
+    bool pass = true;
+
+    if((config.do_verification == 1) || (config.do_verification == 3))
+    {
+        // CPU verification
+        auto ref_gemm    = ReferenceGemmInstance{};
+        auto ref_invoker = ref_gemm.MakeInvoker();
+
+        auto ref_argument = ref_gemm.MakeArgument(
+            a_m_k, b_k_n, c_m_n_host_result, a_element_op, b_element_op, c_element_op);
+
+        std::cout << "Running verification on CPU." << std::endl;
+        ref_invoker.Run(ref_argument);
+
+#ifdef BUILD_INT4_EXAMPLE
+        Tensor<CDataType> c_m_n_device_result_converted(c_m_n_host_result.mDesc);
+
+        c_m_n_device_buf.FromDevice(c_m_n_device_result_converted.mData.data());
+
+        c_m_n_device_result = c_m_n_device_result_converted.CopyAsType<CDataType>();
+
+        return ck::utils::check_err(c_m_n_device_result_converted, c_m_n_host_result);
+#else
+        c_m_n_device_buf.FromDevice(c_m_n_device_result.mData.data());
+
+        pass &= ck::utils::check_err(c_m_n_device_result,
+                                     c_m_n_host_result,
+                                     "Error: Incorrect results!",
+                                     get_rtol<CDataType>(),
+                                     get_atol<CDataType>());
+#endif
+    }
+
+    if((config.do_verification == 2) || (config.do_verification == 3))
+    {
+        // GPU verification
+        auto ref_gemm_gpu    = ReferenceGemmInstanceGPU{};
+        auto ref_invoker_gpu = ref_gemm_gpu.MakeInvoker();
+
+        auto ref_argument_gpu = ref_gemm_gpu.MakeArgument(
+            static_cast<ADataType*>(a_m_k_device_buf.GetDeviceBuffer()),
+            static_cast<BDataType*>(b_k_n_device_buf.GetDeviceBuffer()),
+            static_cast<CDataType*>(c_m_n_device_ref_buf.GetDeviceBuffer()),
+            M,
+            N,
+            K,
+            a_element_op,
+            b_element_op,
+            c_element_op);
+
+        std::cout << "Running verification on GPU." << std::endl;
+        ref_invoker_gpu.Run(ref_argument_gpu, StreamConfig{});
+
+        c_m_n_device_ref_buf.FromDevice(c_m_n_device_ref_result.mData.data());
+        c_m_n_device_buf.FromDevice(c_m_n_device_result.mData.data());
+
+        pass &= ck::utils::check_err(c_m_n_device_result,
+                                     c_m_n_device_ref_result,
+                                     "Error: Incorrect results!",
+                                     get_rtol<CDataType>(),
+                                     get_atol<CDataType>());
+    }
+
+    return pass == true;
+}
+
+bool run_gemm_streamk_example(int argc, char* argv[])
+{
+    ProblemSizeStreamK problem_size;
+    ExecutionConfig config;
+
+    return !parse_cmd_args(argc, argv, problem_size, config) || run_gemm(problem_size, config);
+}
diff --git a/example/30_grouped_conv_fwd_multiple_d/run_grouped_conv_fwd_bias_relu_add_example.inc b/example/30_grouped_conv_fwd_multiple_d/run_grouped_conv_fwd_bias_relu_add_example.inc
index ce42a20be7..627e20e245 100644
--- a/example/30_grouped_conv_fwd_multiple_d/run_grouped_conv_fwd_bias_relu_add_example.inc
+++ b/example/30_grouped_conv_fwd_multiple_d/run_grouped_conv_fwd_bias_relu_add_example.inc
@@ -32,9 +32,9 @@ using BiasLayout = typename LayoutSettingSelector<NDimSpatial>::BiasLayout;
 template <ck::index_t NDimSpatial>
 using ResidualLayout = typename LayoutSettingSelector<NDimSpatial>::ResidualLayout;
 
-#if defined(CK_USE_AMD_MFMA_GFX950)
+// instance for double rate mfma on gfx950 (vs gfx942)
 template <ck::index_t NDimSpatial>
-using DeviceConvFwdInstance =
+using DeviceConvFwdInstance2 =
     ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<
         NDimSpatial,
         InputLayout<NDimSpatial>,
@@ -55,14 +55,14 @@ using DeviceConvFwdInstance =
         1,           //
         256,         // BlockSize
         128,         // MPerBlock
-        256,         // NPerBlock
+        64,          // NPerBlock
         64,          // KPerBlock
         16,          // AK1
         16,          // BK1
         32,          // MPerXdl
         32,          // NPerXdl
         2,           // MXdlPerWave
-        4,           // NXdlPerWave
+        1,           // NXdlPerWave
         S<4, 64, 1>, // ABlockTransferThreadClusterLengths_AK0_M_AK1
         S<1, 0, 2>,  // ABlockTransferThreadClusterArrangeOrder
         S<1, 0, 2>,  // ABlockTransferSrcAccessOrder
@@ -81,7 +81,7 @@ using DeviceConvFwdInstance =
         1,
         S<1, 16, 1, 16>,
         4>;
-#else  // defined(CK_USE_AMD_MFMA_GFX950)
+// instance for gfx942-
 template <ck::index_t NDimSpatial>
 using DeviceConvFwdInstance =
     ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<
@@ -104,14 +104,14 @@ using DeviceConvFwdInstance =
         1,           //
         256,         // BlockSize
         128,         // MPerBlock
-        256,         // NPerBlock
-        16,          // KPerBlock
+        128,         // NPerBlock
+        32,          // KPerBlock
         4,           // AK1
         4,           // BK1
         32,          // MPerXdl
         32,          // NPerXdl
         2,           // MXdlPerWave
-        4,           // NXdlPerWave
+        2,           // NXdlPerWave
         S<4, 64, 1>, // ABlockTransferThreadClusterLengths_AK0_M_AK1
         S<1, 0, 2>,  // ABlockTransferThreadClusterArrangeOrder
         S<1, 0, 2>,  // ABlockTransferSrcAccessOrder
@@ -130,7 +130,6 @@ using DeviceConvFwdInstance =
         1,
         S<1, 16, 1, 16>,
         4>;
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
 
 template <ck::index_t NDimSpatial>
 using HostConvFwdInstance = ck::tensor_operation::host::ReferenceConvFwd<NDimSpatial,
@@ -235,40 +234,67 @@ bool run_grouped_conv_fwd_bias_relu_add(const ExecutionConfig& config,
     copy(conv_param.input_right_pads_, input_right_pads);
 
     // do Conv
-    auto conv    = DeviceConvFwdInstance<NDimSpatial>{};
-    auto invoker = conv.MakeInvoker();
-    auto argument =
-        conv.MakeArgument(in_device_buf.GetDeviceBuffer(),
-                          wei_device_buf.GetDeviceBuffer(),
-                          std::array<const void*, 2>{bias_device_buf.GetDeviceBuffer(),
-                                                     residual_device_buf.GetDeviceBuffer()},
-                          out_device_buf.GetDeviceBuffer(),
-                          a_g_n_c_wis_lengths,
-                          a_g_n_c_wis_strides,
-                          b_g_k_c_xs_lengths,
-                          b_g_k_c_xs_strides,
-                          std::array<std::array<ck::index_t, NDimSpatial + 3>, 2>{
-                              {d0_g_n_k_wos_lengths, d1_g_n_k_wos_lengths}},
-                          std::array<std::array<ck::index_t, NDimSpatial + 3>, 2>{
-                              {d0_g_n_k_wos_strides, d1_g_n_k_wos_strides}},
-                          e_g_n_k_wos_lengths,
-                          e_g_n_k_wos_strides,
-                          conv_filter_strides,
-                          conv_filter_dilations,
-                          input_left_pads,
-                          input_right_pads,
-                          InElementOp{},
-                          WeiElementOp{},
-                          OutElementOp{});
+    using BaseGroupedConvFwdMultipleABD =
+        ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
+            NDimSpatial,
+            InputLayout<NDimSpatial>,
+            WeightLayout<NDimSpatial>,
+            ck::Tuple<BiasLayout<NDimSpatial>, ResidualLayout<NDimSpatial>>,
+            OutputLayout<NDimSpatial>,
+            InKernelDataType,
+            WeiKernelDataType,
+            ck::Tuple<BiasKernelDataType, ResidualKernelDataType>,
+            OutKernelDataType,
+            InElementOp,
+            WeiElementOp,
+            OutElementOp,
+            InKernelDataType,  // AComputeDataType
+            InKernelDataType>; // BComputeDataType
 
-    if(!conv.IsSupportedArgument(argument))
+    static_assert(
+        std::is_base_of<BaseGroupedConvFwdMultipleABD, DeviceConvFwdInstance<NDimSpatial>>::value &&
+        std::is_base_of<BaseGroupedConvFwdMultipleABD, DeviceConvFwdInstance2<NDimSpatial>>::value);
+
+    auto conv  = DeviceConvFwdInstance<NDimSpatial>{};  // instance for gfx942-
+    auto conv2 = DeviceConvFwdInstance2<NDimSpatial>{}; // instance for double rate mfma instruction
+                                                        // on gfx950
+    BaseGroupedConvFwdMultipleABD* op_ptr =
+        (ck::get_device_name() == "gfx950") ? static_cast<BaseGroupedConvFwdMultipleABD*>(&conv2)
+                                            : static_cast<BaseGroupedConvFwdMultipleABD*>(&conv);
+    auto invoker_ptr  = op_ptr->MakeInvokerPointer();
+    auto argument_ptr = op_ptr->MakeArgumentPointer(
+        in_device_buf.GetDeviceBuffer(),
+        wei_device_buf.GetDeviceBuffer(),
+        std::array<const void*, 2>{bias_device_buf.GetDeviceBuffer(),
+                                   residual_device_buf.GetDeviceBuffer()},
+        out_device_buf.GetDeviceBuffer(),
+        a_g_n_c_wis_lengths,
+        a_g_n_c_wis_strides,
+        b_g_k_c_xs_lengths,
+        b_g_k_c_xs_strides,
+        std::array<std::array<ck::index_t, NDimSpatial + 3>, 2>{
+            {d0_g_n_k_wos_lengths, d1_g_n_k_wos_lengths}},
+        std::array<std::array<ck::index_t, NDimSpatial + 3>, 2>{
+            {d0_g_n_k_wos_strides, d1_g_n_k_wos_strides}},
+        e_g_n_k_wos_lengths,
+        e_g_n_k_wos_strides,
+        conv_filter_strides,
+        conv_filter_dilations,
+        input_left_pads,
+        input_right_pads,
+        InElementOp{},
+        WeiElementOp{},
+        OutElementOp{});
+
+    if(!op_ptr->IsSupportedArgument(argument_ptr.get()))
     {
         throw std::runtime_error(
             "wrong! device_conv with the specified compilation parameters does "
             "not support this Conv problem");
     }
 
-    float avg_time = invoker.Run(argument, StreamConfig{nullptr, config.time_kernel});
+    float avg_time =
+        invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, config.time_kernel});
 
     std::size_t flop      = conv_param.GetFlops();
     std::size_t num_btype = conv_param.GetByte<InUserDataType, WeiUserDataType, OutUserDataType>();
@@ -276,7 +302,7 @@ bool run_grouped_conv_fwd_bias_relu_add(const ExecutionConfig& config,
     float tflops     = static_cast<float>(flop) / 1.E9 / avg_time;
     float gb_per_sec = num_btype / 1.E6 / avg_time;
     std::cout << "Perf: " << avg_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
-              << conv.GetTypeString() << std::endl;
+              << op_ptr->GetTypeString() << std::endl;
 
     if(config.do_verification)
     {
diff --git a/include/ck/tensor_operation/gpu/grid/gemm_layernorm/gridwise_gemm_multiple_d_welford_first_half_xdl_cshuffle.hpp b/include/ck/tensor_operation/gpu/grid/gemm_layernorm/gridwise_gemm_multiple_d_welford_first_half_xdl_cshuffle.hpp
index f4d0989088..d728360c55 100644
--- a/include/ck/tensor_operation/gpu/grid/gemm_layernorm/gridwise_gemm_multiple_d_welford_first_half_xdl_cshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gemm_layernorm/gridwise_gemm_multiple_d_welford_first_half_xdl_cshuffle.hpp
@@ -517,8 +517,9 @@ struct GridwiseGemmMultipleDWelfordFirstHalf_xdl_cshuffle
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<ABDataType, half_t>::value || is_same<ABDataType, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<ABDataType, half_t>::value || is_same<ABDataType, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<ABDataType, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack =
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_gemm_xdl_cshuffle_v1.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_gemm_xdl_cshuffle_v1.hpp
index 55e254e015..50b4a734fa 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_gemm_xdl_cshuffle_v1.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_gemm_xdl_cshuffle_v1.hpp
@@ -450,8 +450,9 @@ struct GridwiseBatchedGemmGemm_Xdl_CShuffle
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<FloatAB, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_multiple_d_gemm_multiple_d_xdl_cshuffle_v1.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_multiple_d_gemm_multiple_d_xdl_cshuffle_v1.hpp
index fd16927cc1..79a9410898 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_multiple_d_gemm_multiple_d_xdl_cshuffle_v1.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_multiple_d_gemm_multiple_d_xdl_cshuffle_v1.hpp
@@ -361,9 +361,11 @@ struct GridwiseBatchedGemmMultipleDGemmMultipleD_Xdl_CShuffle
         const auto M = d0_grid_desc_m_n.GetLength(I0);
         const auto N = d0_grid_desc_m_n.GetLength(I1);
 
+        constexpr auto lcm_A0K1_B0K1 = math::lcm(A0K1, B0K1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<A0B0B1DataType, half_t>::value || is_same<A0B0B1DataType, bhalf_t>::value) &&
-             math::lcm(A0K1, B0K1) <= 4)
+            (((is_same<A0B0B1DataType, half_t>::value || is_same<A0B0B1DataType, bhalf_t>::value) &&
+              lcm_A0K1_B0K1 <= 4) ||
+             (is_same<A0B0B1DataType, int8_t>::value && lcm_A0K1_B0K1 <= 8))
                 ? true
                 : false;
         constexpr auto mfma = MfmaSelector<A0B0B1DataType,
@@ -653,8 +655,9 @@ struct GridwiseBatchedGemmMultipleDGemmMultipleD_Xdl_CShuffle
         // sanity check
         constexpr auto lcm_A0K1_B0K1 = math::lcm(A0K1, B0K1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<A0B0B1DataType, half_t>::value || is_same<A0B0B1DataType, bhalf_t>::value) &&
-             lcm_A0K1_B0K1 <= 4)
+            (((is_same<A0B0B1DataType, half_t>::value || is_same<A0B0B1DataType, bhalf_t>::value) &&
+              lcm_A0K1_B0K1 <= 4) ||
+             (is_same<A0B0B1DataType, int8_t>::value && lcm_A0K1_B0K1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack =
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_multiple_d_softmax_gemm_xdl_cshuffle_v1.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_multiple_d_softmax_gemm_xdl_cshuffle_v1.hpp
index 1f7458e68f..d15767f658 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_multiple_d_softmax_gemm_xdl_cshuffle_v1.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_multiple_d_softmax_gemm_xdl_cshuffle_v1.hpp
@@ -343,9 +343,11 @@ struct GridwiseBatchedGemmMultipleDSoftmaxGemm_Xdl_CShuffle
         const auto M = d0_grid_desc_m_n.GetLength(I0);
         const auto N = d0_grid_desc_m_n.GetLength(I1);
 
+        constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
-             math::lcm(AK1, BK1) <= 4)
+            (((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<FloatAB, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr auto mfma =
@@ -560,8 +562,9 @@ struct GridwiseBatchedGemmMultipleDSoftmaxGemm_Xdl_CShuffle
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<FloatAB, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_softmax_gemm_xdl_cshuffle_v1.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_softmax_gemm_xdl_cshuffle_v1.hpp
index f7746b470f..a11d696019 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_softmax_gemm_xdl_cshuffle_v1.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_batched_gemm_softmax_gemm_xdl_cshuffle_v1.hpp
@@ -471,8 +471,9 @@ struct GridwiseBatchedGemmSoftmaxGemm_Xdl_CShuffle
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<FloatAB, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_bias_add_reduce_xdl_cshuffle_v1.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_bias_add_reduce_xdl_cshuffle_v1.hpp
index 8b3f51b9b0..ab97a940a8 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_bias_add_reduce_xdl_cshuffle_v1.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_bias_add_reduce_xdl_cshuffle_v1.hpp
@@ -500,8 +500,9 @@ struct GridwiseGemmBiasAddReduce_k0mk1_k0nk1_mn_xdl_cshuffle_v1
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<FloatAB, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_abd_xdl_cshuffle.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_abd_xdl_cshuffle.hpp
index 344656b13f..79ab3acd92 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_abd_xdl_cshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_abd_xdl_cshuffle.hpp
@@ -674,10 +674,22 @@ struct GridwiseGemmMultipleABD_xdl_cshuffle
         //     c_mtx[MPerBlock, NPerBlock] is distributed among threads, and saved in
         //       register
         // sanity check
-        constexpr index_t KPack = math::max(
-            math::lcm(AK1, BK1),
-            MfmaSelector<AComputeDataType, MPerXdl, NPerXdl, BComputeDataType>::selected_mfma
-                .k_per_blk);
+        constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
+        constexpr bool is_single_rate_mfma =
+            (((is_same<AComputeDataType, half_t>::value ||
+               is_same<AComputeDataType, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<AComputeDataType, int8_t>::value && lcm_AK1_BK1 <= 8))
+                ? true
+                : false;
+
+        constexpr index_t KPack =
+            math::max(lcm_AK1_BK1,
+                      MfmaSelector<AComputeDataType,
+                                   MPerXdl,
+                                   NPerXdl,
+                                   BComputeDataType,
+                                   is_single_rate_mfma>::selected_mfma.k_per_blk);
 
         auto blockwise_gemm = BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_Selector<
             BlockSize,
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp
index 60ee78528d..0e51c6904c 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp
@@ -466,8 +466,9 @@ struct GridwiseGemmMultipleDMultipleR_k0mk1_k0nk1_mn_xdl_cshuffle_v1
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<FloatAB, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp
index 060f6d5d15..d54a00eaa2 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp
@@ -635,10 +635,22 @@ struct GridwiseGemmMultipleD_xdl_cshuffle
         //     c_mtx[MPerBlock, NPerBlock] is distributed among threads, and saved in
         //       register
         // sanity check
-        constexpr index_t KPack = math::max(
-            math::lcm(AK1, BK1),
-            MfmaSelector<AComputeDataType, MPerXdl, NPerXdl, BComputeDataType>::selected_mfma
-                .k_per_blk);
+        constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
+        constexpr bool is_single_rate_mfma =
+            (((is_same<AComputeDataType, half_t>::value ||
+               is_same<AComputeDataType, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<AComputeDataType, int8_t>::value && lcm_AK1_BK1 <= 8))
+                ? true
+                : false;
+
+        constexpr index_t KPack =
+            math::max(lcm_AK1_BK1,
+                      MfmaSelector<AComputeDataType,
+                                   MPerXdl,
+                                   NPerXdl,
+                                   BComputeDataType,
+                                   is_single_rate_mfma>::selected_mfma.k_per_blk);
 
         auto blockwise_gemm = BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_Selector<
             BlockSize,
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle_lds_direct_load.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle_lds_direct_load.hpp
index b4c5d004c4..57b9b02548 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle_lds_direct_load.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle_lds_direct_load.hpp
@@ -600,10 +600,22 @@ struct GridwiseGemmMultipleD_Xdl_CShuffle_LdsDirectLoad
         //     b_mtx[K0PerBlock, NPerBlock] is in LDS
         //     c_mtx[MPerBlock, NPerBlock] is distributed among threads, and saved in
         //       register
-        constexpr index_t KPack = math::max(
-            math::lcm(AK1, BK1),
-            MfmaSelector<AComputeDataType, MPerXdl, NPerXdl, BComputeDataType>::selected_mfma
-                .k_per_blk);
+        constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
+        constexpr bool is_single_rate_mfma =
+            (((is_same<AComputeDataType, half_t>::value ||
+               is_same<AComputeDataType, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<AComputeDataType, int8_t>::value && lcm_AK1_BK1 <= 8))
+                ? true
+                : false;
+
+        constexpr index_t KPack =
+            math::max(lcm_AK1_BK1,
+                      MfmaSelector<AComputeDataType,
+                                   MPerXdl,
+                                   NPerXdl,
+                                   BComputeDataType,
+                                   is_single_rate_mfma>::selected_mfma.k_per_blk);
 
         auto blockwise_gemm = BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_Selector<
             BlockSize,
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_splitk_cshuffle.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_splitk_cshuffle.hpp
index d1d97da5b0..88d6be234c 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_splitk_cshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_splitk_cshuffle.hpp
@@ -601,8 +601,9 @@ struct GridwiseGemmMultipleD_xdl_splitk_cshuffle
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<AComputeType, half_t>::value || is_same<AComputeType, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<AComputeType, half_t>::value || is_same<AComputeType, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<AComputeType, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_reduce_xdl_cshuffle_v1.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_reduce_xdl_cshuffle_v1.hpp
index 7105fa7012..56581256dc 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_reduce_xdl_cshuffle_v1.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_reduce_xdl_cshuffle_v1.hpp
@@ -453,8 +453,9 @@ struct GridwiseGemmReduce_k0mk1_k0nk1_mn_xdl_cshuffle_v1
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<FloatAB, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_split_k_multiple_d_xdl_cshuffle.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_split_k_multiple_d_xdl_cshuffle.hpp
index 3429c20e73..23b4aec3b0 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_split_k_multiple_d_xdl_cshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_split_k_multiple_d_xdl_cshuffle.hpp
@@ -583,8 +583,9 @@ struct GridwiseGemmSplitKMultipleD_xdl_cshuffle
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<ABDataType, half_t>::value || is_same<ABDataType, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<ABDataType, half_t>::value || is_same<ABDataType, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<ABDataType, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack =
@@ -1015,8 +1016,9 @@ struct GridwiseGemmSplitKMultipleD_xdl_cshuffle
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<ABDataType, half_t>::value || is_same<ABDataType, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<ABDataType, half_t>::value || is_same<ABDataType, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<ABDataType, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack =
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_split_k_multiple_d_xdl_cshuffle_v2.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_split_k_multiple_d_xdl_cshuffle_v2.hpp
index d7c87a170c..44c1e936bd 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_split_k_multiple_d_xdl_cshuffle_v2.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_split_k_multiple_d_xdl_cshuffle_v2.hpp
@@ -597,8 +597,9 @@ struct GridwiseGemmMultipleD_xdl_splitk_cshuffle
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<ComputeType, half_t>::value || is_same<ComputeType, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<ComputeType, half_t>::value || is_same<ComputeType, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<ComputeType, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_bwd_weight_v3.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_bwd_weight_v3.hpp
index 08d9386d72..4f5fedcd83 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_bwd_weight_v3.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_bwd_weight_v3.hpp
@@ -81,8 +81,9 @@ struct GridwiseGemm_xdl_cshuffle_v3
 
     static constexpr auto lcm_AK1_BK1 = math::lcm(AK1Number, BK1Number);
     static constexpr bool is_single_rate_mfma =
-        ((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
-         lcm_AK1_BK1 <= 4)
+        (((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
+          lcm_AK1_BK1 <= 4) ||
+         (is_same<ComputeTypeA, int8_t>::value && lcm_AK1_BK1 <= 8))
             ? true
             : false;
     static constexpr index_t KPack =
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_streamk_v3.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_streamk_v3.hpp
index fcb12f4a14..cc8ae1806a 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_streamk_v3.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_streamk_v3.hpp
@@ -141,8 +141,9 @@ struct GridwiseGemm_xdl_cshuffle_streamk_v3
 
     static constexpr auto lcm_AK1_BK1 = math::lcm(AK1Number, BK1Number);
     static constexpr bool is_single_rate_mfma =
-        ((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
-         lcm_AK1_BK1 <= 4)
+        (((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
+          lcm_AK1_BK1 <= 4) ||
+         (is_same<ComputeTypeA, int8_t>::value && lcm_AK1_BK1 <= 8))
             ? true
             : false;
     static constexpr index_t KPack =
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v1.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v1.hpp
index c293d64ef0..240bc464e1 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v1.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v1.hpp
@@ -810,9 +810,17 @@ struct GridwiseGemm_k0mk1_k0nk1_mn_xdl_cshuffle_v1
         //     c_mtx[MPerBlock, NPerBlock] is distributed among threads, and saved in
         //       register
         // sanity check
+        constexpr auto lcm_AK1_BK1 = math::lcm(AK1Number, BK1Number);
+        constexpr bool is_single_rate_mfma =
+            (((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<ComputeTypeA, int8_t>::value && lcm_AK1_BK1 <= 8))
+                ? true
+                : false;
         constexpr index_t KPack = math::max(
-            math::lcm(AK1Number, BK1Number),
-            MfmaSelector<ComputeTypeA, MPerXdl, NPerXdl, ComputeTypeB>::selected_mfma.k_per_blk);
+            lcm_AK1_BK1,
+            MfmaSelector<ComputeTypeA, MPerXdl, NPerXdl, ComputeTypeB, is_single_rate_mfma>::
+                selected_mfma.k_per_blk);
 
         auto blockwise_gemm = BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_Selector<
             BlockSize,
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v2.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v2.hpp
index af91721c8a..c7d44e842d 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v2.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v2.hpp
@@ -871,8 +871,9 @@ struct GridwiseGemm_xdl_cshuffle_v2
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1Number, BK1Number);
         constexpr bool is_single_rate_mfma =
-            ((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<ComputeTypeA, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3.hpp
index d4c915aa5e..55639f4aee 100755
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3.hpp
@@ -149,8 +149,9 @@ struct GridwiseGemm_xdl_cshuffle_v3
 
     static constexpr auto lcm_AK1_BK1 = math::lcm(AK1Number, BK1Number);
     static constexpr bool is_single_rate_mfma =
-        ((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
-         lcm_AK1_BK1 <= 4)
+        (((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
+          lcm_AK1_BK1 <= 4) ||
+         (is_same<ComputeTypeA, int8_t>::value && lcm_AK1_BK1 <= 8))
             ? true
             : false;
     static constexpr index_t KPack =
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_b_scale.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_b_scale.hpp
index 2e62110416..27818b6964 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_b_scale.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_b_scale.hpp
@@ -157,8 +157,9 @@ struct GridwiseGemm_xdl_cshuffle_v3
 
     static constexpr auto lcm_AK1_BK1 = math::lcm(AK1Number, BK1Number);
     static constexpr bool is_single_rate_mfma =
-        ((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
-         lcm_AK1_BK1 <= 4)
+        (((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
+          lcm_AK1_BK1 <= 4) ||
+         (is_same<ComputeTypeA, int8_t>::value && lcm_AK1_BK1 <= 8))
             ? true
             : false;
     static constexpr index_t KPack =
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_abd.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_abd.hpp
index f9071bd29d..b805f600d5 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_abd.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_abd.hpp
@@ -193,9 +193,17 @@ struct GridwiseGemm_xdl_cshuffle_v3
     using BsGridPointer = decltype(MakeBsGridPointer());
     using DsGridPointer = decltype(MakeDsGridPointer());
 
-    static constexpr index_t KPack = math::max(
-        math::lcm(AK1Number, BK1Number),
-        MfmaSelector<ComputeTypeA, MPerXdl, NPerXdl, ComputeTypeB>::selected_mfma.k_per_blk);
+    static constexpr auto lcm_AK1_BK1 = math::lcm(AK1Number, BK1Number);
+    static constexpr bool is_single_rate_mfma =
+        (((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
+          lcm_AK1_BK1 <= 4) ||
+         (is_same<ComputeTypeA, int8_t>::value && lcm_AK1_BK1 <= 8))
+            ? true
+            : false;
+    static constexpr index_t KPack =
+        math::max(lcm_AK1_BK1,
+                  MfmaSelector<ComputeTypeA, MPerXdl, NPerXdl, ComputeTypeB, is_single_rate_mfma>::
+                      selected_mfma.k_per_blk);
 
     using ThisThreadBlock = ThisThreadBlock<BlockSize>;
 
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d.hpp
index a9e73bf461..4163d1d01a 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d.hpp
@@ -179,9 +179,18 @@ struct GridwiseGemmMultiD_xdl_cshuffle_v3
 
     using DsGridPointer = decltype(MakeDsGridPointer());
 
-    static constexpr index_t KPack = math::max(
-        math::lcm(AK1Number, BK1Number),
-        MfmaSelector<ComputeTypeA, MPerXdl, NPerXdl, ComputeTypeB>::selected_mfma.k_per_blk);
+    static constexpr auto lcm_AK1_BK1 = math::lcm(AK1Number, BK1Number);
+    static constexpr bool is_single_rate_mfma =
+        (((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
+          lcm_AK1_BK1 <= 4) ||
+         (is_same<ComputeTypeA, int8_t>::value && lcm_AK1_BK1 <= 8))
+            ? true
+            : false;
+
+    static constexpr index_t KPack =
+        math::max(lcm_AK1_BK1,
+                  MfmaSelector<ComputeTypeA, MPerXdl, NPerXdl, ComputeTypeB, is_single_rate_mfma>::
+                      selected_mfma.k_per_blk);
 
     using ThisThreadBlock = ThisThreadBlock<BlockSize>;
 
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d_ab_scale.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d_ab_scale.hpp
index 813acfa656..d10db3225e 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d_ab_scale.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v3_multi_d_ab_scale.hpp
@@ -149,9 +149,18 @@ struct GridwiseGemmMultiD_ABScale_xdl_cshuffle_v3
 
     using DsGridPointer = decltype(MakeDsGridPointer());
 
-    static constexpr index_t KPack = math::max(
-        math::lcm(AK1Number, BK1Number),
-        MfmaSelector<ComputeTypeA, MPerXdl, NPerXdl, ComputeTypeB>::selected_mfma.k_per_blk);
+    static constexpr auto lcm_AK1_BK1 = math::lcm(AK1Number, BK1Number);
+    static constexpr bool is_single_rate_mfma =
+        (((is_same<ComputeTypeA, half_t>::value || is_same<ComputeTypeA, bhalf_t>::value) &&
+          lcm_AK1_BK1 <= 4) ||
+         (is_same<ComputeTypeA, int8_t>::value && lcm_AK1_BK1 <= 8))
+            ? true
+            : false;
+
+    static constexpr index_t KPack =
+        math::max(lcm_AK1_BK1,
+                  MfmaSelector<ComputeTypeA, MPerXdl, NPerXdl, ComputeTypeB, is_single_rate_mfma>::
+                      selected_mfma.k_per_blk);
 
     using ThisThreadBlock = ThisThreadBlock<BlockSize>;
 
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_layernorm_cshuffle_v1.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_layernorm_cshuffle_v1.hpp
index 0a62464cc2..b435fd5d5a 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_layernorm_cshuffle_v1.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_layernorm_cshuffle_v1.hpp
@@ -491,8 +491,9 @@ struct GridwiseGemmLayernorm_k0mk1_k0nk1_mn_xdl_cshuffle_v1
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<FloatAB, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_waveletmodel_cshuffle.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_waveletmodel_cshuffle.hpp
index 6a4b1cc14b..ad65e75ef9 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_waveletmodel_cshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_waveletmodel_cshuffle.hpp
@@ -489,8 +489,9 @@ struct GridwiseGemm_k0mk1_k0nk1_mn_xdl_waveletmodel_cshuffle
             // branch early for math wave
             constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
             constexpr bool is_single_rate_mfma =
-                ((is_same<ABDataType, half_t>::value || is_same<ABDataType, bhalf_t>::value) &&
-                 lcm_AK1_BK1 <= 4)
+                (((is_same<ABDataType, half_t>::value || is_same<ABDataType, bhalf_t>::value) &&
+                  lcm_AK1_BK1 <= 4) ||
+                 (is_same<ABDataType, int8_t>::value && lcm_AK1_BK1 <= 8))
                     ? true
                     : false;
             constexpr index_t KPack = math::max(
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_bwd_weight.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_bwd_weight.hpp
index b41e747a3a..168c553180 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_bwd_weight.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_bwd_weight.hpp
@@ -741,11 +741,17 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_bwd_weight
         //     c_mtx[MPerBlock, NPerBlock] is distributed among threads, and saved in
         //       register
         // sanity check
+        constexpr bool is_single_rate_mfma =
+            (((is_same<FloatAAdjusted, half_t>::value || is_same<FloatAAdjusted, bhalf_t>::value) &&
+              K1 <= 4) ||
+             (is_same<FloatAAdjusted, int8_t>::value && K1 <= 8))
+                ? true
+                : false;
 
-        constexpr index_t KPack =
-            math::max(K1,
-                      MfmaSelector<FloatAAdjusted, MPerXDL, NPerXDL, FloatBAdjusted>::selected_mfma
-                          .k_per_blk);
+        constexpr index_t KPack = math::max(
+            K1,
+            MfmaSelector<FloatAAdjusted, MPerXDL, NPerXDL, FloatBAdjusted, is_single_rate_mfma>::
+                selected_mfma.k_per_blk);
 
         auto blockwise_gemm =
             BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1<BlockSize,
diff --git a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v3r1.hpp b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v3r1.hpp
index 7db8798695..f433b3c974 100644
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v3r1.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v3r1.hpp
@@ -448,8 +448,9 @@ struct GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v3r1
         // sanity check
         constexpr auto lcm_AK1_BK1 = math::lcm(AK1, BK1);
         constexpr bool is_single_rate_mfma =
-            ((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
-             lcm_AK1_BK1 <= 4)
+            (((is_same<FloatAB, half_t>::value || is_same<FloatAB, bhalf_t>::value) &&
+              lcm_AK1_BK1 <= 4) ||
+             (is_same<FloatAB, int8_t>::value && lcm_AK1_BK1 <= 8))
                 ? true
                 : false;
         constexpr index_t k_pack = math::max(
diff --git a/include/ck/tensor_operation/gpu/warp/xdlops_gemm.hpp b/include/ck/tensor_operation/gpu/warp/xdlops_gemm.hpp
index 8c0b950941..9f6c0b5648 100644
--- a/include/ck/tensor_operation/gpu/warp/xdlops_gemm.hpp
+++ b/include/ck/tensor_operation/gpu/warp/xdlops_gemm.hpp
@@ -1053,40 +1053,49 @@ struct MfmaSelector
 #endif
     }
 
+    template <>
+    constexpr auto GetMfma<int8_t, 32, 32, int8_t, false>()
+    {
 #if defined(__gfx950__)
-    template <>
-    constexpr auto GetMfma<int8_t, 32, 32>()
-    {
         return MfmaInstr::mfma_i32_32x32x32i8;
-    }
-    template <>
-    constexpr auto GetMfma<int8_t, 16, 16>()
-    {
-        return MfmaInstr::mfma_i32_16x16x64i8;
-    }
 #elif defined(__gfx942__)
-    template <>
-    constexpr auto GetMfma<int8_t, 32, 32>()
-    {
         return MfmaInstr::mfma_i32_32x32x16i8;
-    }
-    template <>
-    constexpr auto GetMfma<int8_t, 16, 16>()
-    {
-        return MfmaInstr::mfma_i32_16x16x32i8;
-    }
 #else
-    template <>
-    constexpr auto GetMfma<int8_t, 32, 32>()
-    {
         return MfmaInstr::mfma_i32_32x32x8i8;
-    }
-    template <>
-    constexpr auto GetMfma<int8_t, 16, 16>()
-    {
-        return MfmaInstr::mfma_i32_16x16x16i8;
-    }
 #endif
+    }
+
+    template <>
+    constexpr auto GetMfma<int8_t, 32, 32, int8_t, true>()
+    {
+#if defined(__gfx942__) || defined(__gfx950__)
+        return MfmaInstr::mfma_i32_32x32x16i8;
+#else
+        return MfmaInstr::mfma_i32_32x32x8i8;
+#endif
+    }
+
+    template <>
+    constexpr auto GetMfma<int8_t, 16, 16, int8_t, false>()
+    {
+#if defined(__gfx950__)
+        return MfmaInstr::mfma_i32_16x16x64i8;
+#elif defined(__gfx942__)
+        return MfmaInstr::mfma_i32_16x16x32i8;
+#else
+        return MfmaInstr::mfma_i32_16x16x16i8;
+#endif
+    }
+
+    template <>
+    constexpr auto GetMfma<int8_t, 16, 16, int8_t, true>()
+    {
+#if defined(__gfx942__) || defined(__gfx950__)
+        return MfmaInstr::mfma_i32_16x16x32i8;
+#else
+        return MfmaInstr::mfma_i32_16x16x16i8;
+#endif
+    }
 
     template <>
     constexpr auto GetMfma<f8_t, 32, 32>()
@@ -1440,12 +1449,13 @@ struct XdlopsGemm
     }
 
     // Falls back to single rate instruction on gfx950 if KPack <= 4; no change on gfx942-
-    static constexpr auto
-        mfma = MfmaSelector < base_type,
-        MPerXdlops, NPerXdlops, additional_type,
-        ((is_same<base_type, half_t>::value || is_same<base_type, bhalf_t>::value) && KPack <= 4)
-            ? true
-            : false > {};
+    static constexpr auto mfma = MfmaSelector < base_type, MPerXdlops, NPerXdlops, additional_type,
+                          (((is_same<base_type, half_t>::value ||
+                             is_same<base_type, bhalf_t>::value) &&
+                            KPack <= 4) ||
+                           (is_same<base_type, int8_t>::value && KPack <= 8))
+                              ? true
+                              : false > {};
 
     static constexpr auto mfma_instr = mfma.selected_mfma;
 
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_two_stage_xdl_instance.hpp b/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_two_stage_xdl_instance.hpp
index 3ebfd0bb7d..bea22da2c2 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_two_stage_xdl_instance.hpp
+++ b/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_two_stage_xdl_instance.hpp
@@ -41,13 +41,11 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_f16_generic_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>
         // clang-format on
         >;
@@ -60,13 +58,11 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineScheduler Scheduler,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_f16_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
 
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              2,              2,      false,  S<4, 16,  1>, S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 2>,
@@ -110,13 +106,11 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_bf16_generic_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>
         // clang-format on
         >;
@@ -129,13 +123,11 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineScheduler Scheduler,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_nhwgc_xdl_c_shuffle_bf16_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1>,
 
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              2,              2,      false,  S<4, 16,  1>, S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 2>,
@@ -179,13 +171,11 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_ngchw_xdl_c_shuffle_f16_generic_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1, F16, F16, 1, 1>
         // clang-format on
         >;
@@ -199,13 +189,11 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineScheduler Scheduler,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_ngchw_xdl_c_shuffle_f16_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1, F16, F16, 1, 1>,
     
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              2,              2,      false,  S<4, 16,  1>, S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 2, F16, F16, 2, 2>,
@@ -241,13 +229,11 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_ngchw_xdl_c_shuffle_bf16_generic_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,    F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1, BF16, BF16, 1, 1>
         // clang-format on
         >;
@@ -260,13 +246,11 @@ template <ck::index_t NDimSpatial,
           BlockGemmPipelineScheduler Scheduler,
           BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_two_stage_ngchw_xdl_c_shuffle_bf16_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
         //#########################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
         //#########################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
         //#########################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,    F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              1,              4,      false,  S<4, 8,  1>,  S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 1, BF16, BF16, 1, 1>,
     
         DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,    F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>, S<2, 0, 1>,  S<1, 0, 2>,                   1,              2,              2,      false,  S<4, 16,  1>, S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,                  1, Scheduler, PipelineVersion, 2, BF16, BF16, 2, 2>,
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp b/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp
index f9b3b43795..e7bbf8a26a 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp
+++ b/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp
@@ -49,6 +49,23 @@ static constexpr auto GemmMNKPadding = GemmSpecialization::MNKPadding;
 static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
 static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
+// double rate mfma instances on gfx950
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_bf16_comp_instances_2x = std::tuple<
+    // clang-format off
+        //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,  16,  16,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
+    // clang-format on
+    >;
+
 template <index_t NDimSpatial,
           typename ALayout,
           typename BLayout,
@@ -56,14 +73,11 @@ template <index_t NDimSpatial,
           typename ELayout,
           ConvolutionForwardSpecialization ConvSpec>
 using device_grouped_conv_fwd_xdl_bf16_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
         //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
         //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,  16,  16,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
-#else // defined(CK_USE_AMD_MFMA_GFX950)
         // Compute friendly
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
@@ -71,18 +85,45 @@ using device_grouped_conv_fwd_xdl_bf16_comp_instances = std::tuple<
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-        // AGPR Spill when use permuted lds layout. so, use padding for these two.
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 64, 1, 4>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   256,    32,   8,   8,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   128,    32,   8,   8,  32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,    64,    64,   8,   8,  32,   32,    2,    1,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,    64,   128,    64,   8,   8,  32,   32,    1,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,    64,    64,    64,   8,   8,  32,   32,    1,    1,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
+    // clang-format on
+    >;
+
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_bf16_comp_instances_part2 = std::tuple<
+    // clang-format off
+        // AGPR Spill when use permuted lds layout. so, use padding for these two.
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 64, 1, 4>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>
+    // clang-format on
+    >;
+
+// double rate mfma instances on gfx950
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_f16_comp_instances_2x = std::tuple<
+    // clang-format off
+        //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,  16,  16,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
 
@@ -93,16 +134,24 @@ template <index_t NDimSpatial,
           typename ELayout,
           ConvolutionForwardSpecialization ConvSpec>
 using device_grouped_conv_fwd_xdl_f16_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
         //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
         //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,  16,  16,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
-#else // defined(CK_USE_AMD_MFMA_GFX950)
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>
+    // clang-format on
+    >;
+
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_f16_comp_instances_part2 = std::tuple<
+    // clang-format off
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    32,   8,   8,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
@@ -115,7 +164,6 @@ using device_grouped_conv_fwd_xdl_f16_comp_instances = std::tuple<
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   256,    32,   8,   8,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   128,    32,   8,   8,  32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
     // clang-format on
     >;
 
@@ -138,6 +186,23 @@ using device_grouped_conv_fwd_xdl_f32_comp_instances = std::tuple<
     // clang-format on
     >;
 
+// double rate mfma instances on gfx950
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_int8_comp_instances_2x = std::tuple<
+    // clang-format off
+        //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,DsLayout,ELayout,int8_t,int8_t,int32_t,   int8_t,    DsLayout,int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,             256,   128,   128,   128,  32,  32,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,              16,        1,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,         1,            1,           1,              S<1, 64, 1, 4>,              16,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
+    // clang-format on
+    >;
+
 template <index_t NDimSpatial,
           typename ALayout,
           typename BLayout,
@@ -145,16 +210,24 @@ template <index_t NDimSpatial,
           typename ELayout,
           ConvolutionForwardSpecialization ConvSpec>
 using device_grouped_conv_fwd_xdl_int8_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
         //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
         //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,DsLayout,ELayout,int8_t,int8_t,int32_t,   int8_t,    DsLayout,int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,             256,   128,   128,   128,  32,  32,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,              16,        1,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,         1,            1,           1,              S<1, 64, 1, 4>,              16,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
-#else // defined(CK_USE_AMD_MFMA_GFX950)
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    32,   8,   8,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>
+    // clang-format on
+    >;
+
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_int8_comp_instances_part2 = std::tuple<
+    // clang-format off
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    32,   8,   8,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   256,    32,   8,   8,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
@@ -163,10 +236,8 @@ using device_grouped_conv_fwd_xdl_int8_comp_instances = std::tuple<
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   256,    32,   8,   8,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   256,   128,    32,   8,   8,  32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   int8_t,   int8_t,     int32_t,      int8_t,    DsLayout,   int8_t, PassThrough, PassThrough, PassThrough,       ConvSpec, GemmMNKPadding,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
     // clang-format on
     >;
-
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_merged_groups_instance.hpp b/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_merged_groups_instance.hpp
index 9114d5c1fb..153cc61b09 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_merged_groups_instance.hpp
+++ b/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_merged_groups_instance.hpp
@@ -40,21 +40,34 @@ template <index_t NDimSpatial,
           typename ELayout,
           ConvolutionForwardSpecialization ConvSpec>
 using device_grouped_conv_fwd_xdl_merged_groups_bf16_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer| ACompute| BCompute| BlockGemm| NumGroups|
         //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|     Type|     Type|  Pipeline|   ToMerge|
         //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|         |         | Scheduler|          |
         //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |         |         |          |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 8>,
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 16>,
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 32>
-#else // defined(CK_USE_AMD_MFMA_GFX950)
         // Instances with NumGroupsPerBatch > 1
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 8>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 16>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 32>
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
+    // clang-format on
+    >;
+
+// double rate mfma instances on gfx950
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_merged_groups_bf16_instances_2x = std::tuple<
+    // clang-format off
+        //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer| ACompute| BCompute| BlockGemm| NumGroups|
+        //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|     Type|     Type|  Pipeline|   ToMerge|
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|         |         | Scheduler|          |
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |         |         |          |          |
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 16>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,  BF16,  BF16,     F32,     BF16,    DsLayout,  BF16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, BF16, BF16, LoopScheduler::Default, 32>
     // clang-format on
     >;
 
@@ -65,22 +78,35 @@ template <index_t NDimSpatial,
           typename ELayout,
           ConvolutionForwardSpecialization ConvSpec>
 using device_grouped_conv_fwd_xdl_merged_groups_f16_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
         //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
         //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        // Instances with NumGroupsPerBatch > 1
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 8>,
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 16>,
-        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 32>
-#else // defined(CK_USE_AMD_MFMA_GFX950)
         // Instances with NumGroupsPerBatch > 1
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 8>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 16>,
         DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     16,   4, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 32>
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
+    // clang-format on
+    >;
+
+// double rate mfma instances on gfx950
+template <index_t NDimSpatial,
+          typename ALayout,
+          typename BLayout,
+          typename DsLayout,
+          typename ELayout,
+          ConvolutionForwardSpecialization ConvSpec>
+using device_grouped_conv_fwd_xdl_merged_groups_f16_instances_2x = std::tuple<
+    // clang-format off
+        //########################################|     NumDim|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|          Ds| EData|           A|           B|         CDE|    ConvForward|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //########################################|    Spatial| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|    DataType|  Type| Elementwise| Elementwise| Elementwise| Specialization| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //########################################|           |       |       |            |       |      |      |        |         |            |      |            |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        // Instances with NumGroupsPerBatch > 1
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 8>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 16>,
+        DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<NDimSpatial,ALayout,BLayout,    DsLayout,ELayout,   F16,   F16,     F32,      F16,    DsLayout,   F16, PassThrough, PassThrough, PassThrough,                  ConvSpec, GemmMNKPadding,  1,  64,    64,    16,     32,   8, 4,  16,   16,    4,    1,  S< 4, 16,  1>, S<0, 2, 1>,     S<0, 2, 1>,                   1,              4,              4,      1,  S< 4, 16,  1>,   S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              4,      1,           1,           1,   S<1, 16, 1, 4>,                  1, F16, F16, LoopScheduler::Default, 32>
     // clang-format on
     >;
 
diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instance.cpp
index dad67b396f..f3a32bab54 100644
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instance.cpp
@@ -8,6 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_batched_gemm_xdl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -27,15 +28,12 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 // Compilation parameters for a[k, m] * b[k, n] = c[m, n]
 using device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##########|        AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
         //##########|         Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
         //##########|             |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
         //##########|             |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -70,6 +68,17 @@ using device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instances = std::tuple<
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instances_2x = std::tuple<
+    // clang-format off
+        //##########|        AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
+        //##########|         Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
+        //##########|             |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
+        //##########|             |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Col,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 void add_device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instances(
     std::vector<std::unique_ptr<
@@ -78,6 +87,12 @@ void add_device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instances(
 {
     add_device_operation_instances(instances,
                                    device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instance.cpp
index 62df9a81b5..43810d9a5e 100644
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instance.cpp
@@ -8,6 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_batched_gemm_xdl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -27,15 +28,12 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 // Compilation parameters for a[k, m] * b[n, k] = c[m, n]
 using device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##########|        AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
         //##########|         Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
         //##########|             |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
         //##########|             |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -70,6 +68,17 @@ using device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instances = std::tuple<
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instances_2x = std::tuple<
+    // clang-format off
+        //##########|        AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
+        //##########|         Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
+        //##########|             |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
+        //##########|             |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 void add_device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instances(
     std::vector<std::unique_ptr<
@@ -78,6 +87,12 @@ void add_device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instances(
 {
     add_device_operation_instances(instances,
                                    device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instance.cpp
index 23e176f6e8..b3bf65325b 100644
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instance.cpp
@@ -8,6 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_batched_gemm_xdl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -26,29 +27,33 @@ using S = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 using device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_generic_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#################| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
         //#################|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
         //#################|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
         //#################|      |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,    64,    16,    16,     4, 16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,    64,    16,    16,     4,  8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_generic_instances_2x = std::tuple<
+    // clang-format off
+        //#################| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
+        //#################|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
+        //#################|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
+        //#################|      |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
+        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,    64,    16,    16,     4, 16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 // Compilation parameters for a[m, k] * b[k, n] = c[m, n]
 using device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#################| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
         //#################|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
         //#################|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
         //#################|      |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -110,6 +115,17 @@ using device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instances = std::tuple<
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instances_2x = std::tuple<
+    // clang-format off
+        //#################| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
+        //#################|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
+        //#################|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
+        //#################|      |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 void add_device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instances(
     std::vector<std::unique_ptr<
@@ -120,6 +136,14 @@ void add_device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instances(
         instances, device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_generic_instances{});
     add_device_operation_instances(instances,
                                    device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_generic_instances_2x{});
+        add_device_operation_instances(
+            instances, device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instance.cpp
index 0216b002ac..39580485ec 100644
--- a/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm/device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instance.cpp
@@ -8,6 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_batched_gemm_xdl.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -26,29 +27,33 @@ using S = ck::Sequence<Is...>;
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
 using device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_generic_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#################| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
         //#################|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
         //#################|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
         //#################|      |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    32,    64,     4, 16,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_generic_instances_2x = std::tuple<
+    // clang-format off
+        //#################| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
+        //#################|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
+        //#################|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
+        //#################|      |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
+        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,    64,    32,    64,     4, 16,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
 using device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#################| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
         //#################|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
         //#################|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
         //#################|      |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -98,6 +103,17 @@ using device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instances = std::tuple<
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instances_2x = std::tuple<
+    // clang-format off
+        //#################| AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CThreadTransfer| CThreadTransfer| NumGemmK|          LoopScheduler|                    Pipeline|
+        //#################|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar| Prefetch|                       |                            |
+        //#################|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector| Stage   |                       |                            |
+        //#################|      |      |      |        |        |        |        |            |            |            |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |         |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceBatchedGemmXdl<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,        1, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 void add_device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instances(
     std::vector<std::unique_ptr<
@@ -108,6 +124,14 @@ void add_device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instances(
         instances, device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_generic_instances{});
     add_device_operation_instances(instances,
                                    device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_generic_instances_2x{});
+        add_device_operation_instances(
+            instances, device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm/device_batched_gemm_softmax_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm/device_batched_gemm_softmax_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
index 6af3c2654e..cf23d01bf2 100644
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm/device_batched_gemm_softmax_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm/device_batched_gemm_softmax_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
@@ -33,13 +33,11 @@ static constexpr auto GemmPadded  = ck::tensor_operation::device::GemmSpecializa
 template <bool Masking>
 using device_batched_gemm_softmax_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //#######################################| ALayout| B0Layout| B1Layout| CLayout| AData| B0Data| B1Data| CData| AccData| CShuffle|           A|          B0|        Acc0|          B1|           C|           GEMM| NumGemmK| Block| Gemm01| Gemm0| Gemm0| Gemm1| Gemm1| AK1| BK1| B1K1| MPer| NPer| Gemm0| Gemm0| Gemm1|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockLds|  B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|    MaskOut|
         //#######################################|        |         |         |        |  Type|   Type|   Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size|   MPer|  NPer|  KPer|  NPer|  KPer|    |    |     |  XDL|  XDL|  MXdl|  NXdl|  NXdl|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|    ThreadCluster|   ThreadCluster|  SrcAccessOrder|    SrcVectorDim|       SrcScalar|       DstScalar|  AddExtraN|    ThreadCluster|   ThreadCluster|  SrcAccessOrder|    SrcVectorDim|       SrcScalar|       DstScalar|  AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|      Upper|
         //#######################################|        |         |         |        |      |       |       |      |        |         |   Operation|   Operation|   Operation|   Operation|   Operation|               |    Stage|      |  Block| Block| Block| Block| Block|    |    |     |     |     |   Per|   Per|   Per| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          |  Lengths_K0_N_K1|    ArrangeOrder|                |                |       PerVector|    PerVector_K1|           |  Lengths_K0_N_K1|    ArrangeOrder|                |                |       PerVector|    PerVector_K1|           |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|   Triangle|
         //#######################################|        |         |         |        |      |       |       |      |        |         |            |            |            |            |            |               |         |      |       |      |      |      |      |    |    |     |     |     |  Wave|  Wave|  Wave|                |               |               |               |               |               |          |                 |                |                |                |                |                |           |                 |                |                |                |                |                |           |            |            |                             |                |           |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle<     Row,      Col,      Row,     Row,   F16,    F16,    F16,   F16,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault,        1,   256,    256,   128,    32,    64,    32,   8,   8,    2,   32,   32,     2,     4,     2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S<16, 16, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8,   Masking>,
         DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle<     Row,      Col,      Row,     Row,   F16,    F16,    F16,   F16,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault,        1,   256,    256,   128,    32,   128,    32,   8,   8,    2,   32,   32,     2,     4,     4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8,   Masking>,
         DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle<     Row,      Col,      Row,     Row,   F16,    F16,    F16,   F16,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault,        1,   256,    128,   256,    32,    64,    32,   8,   8,    2,   32,   32,     1,     8,     2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S<16, 16, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8,   Masking>,
@@ -61,13 +59,11 @@ using device_batched_gemm_softmax_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_
 template <bool Masking>
 using device_batched_gemm_softmax_gemm_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_irregular_k_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //#######################################| ALayout| B0Layout| B1Layout| CLayout| AData| B0Data| B1Data| CData| AccData| CShuffle|           A|          B0|        Acc0|          B1|           C|           GEMM| NumGemmK| Block| Gemm01| Gemm0| Gemm0| Gemm1| Gemm1| AK1| BK1| B1K1| MPer| NPer| Gemm0| Gemm0| Gemm1|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockLds|  B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|    MaskOut|
         //#######################################|        |         |         |        |  Type|   Type|   Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size|   MPer|  NPer|  KPer|  NPer|  KPer|    |    |     |  XDL|  XDL|  MXdl|  NXdl|  NXdl|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|    ThreadCluster|   ThreadCluster|  SrcAccessOrder|    SrcVectorDim|       SrcScalar|       DstScalar|  AddExtraN|    ThreadCluster|   ThreadCluster|  SrcAccessOrder|    SrcVectorDim|       SrcScalar|       DstScalar|  AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|      Upper|
         //#######################################|        |         |         |        |      |       |       |      |        |         |   Operation|   Operation|   Operation|   Operation|   Operation|               |    Stage|      |  Block| Block| Block| Block| Block|    |    |     |     |     |   Per|   Per|   Per| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          |  Lengths_K0_N_K1|    ArrangeOrder|                |                |       PerVector|    PerVector_K1|           |  Lengths_K0_N_K1|    ArrangeOrder|                |                |       PerVector|    PerVector_K1|           |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|   Triangle|
         //#######################################|        |         |         |        |      |       |       |      |        |         |            |            |            |            |            |               |         |      |       |      |      |      |      |    |    |     |     |     |  Wave|  Wave|  Wave|                |               |               |               |               |               |          |                 |                |                |                |                |                |           |                 |                |                |                |                |                |           |            |            |                             |                |           |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle<     Row,      Col,      Row,     Row,   F16,    F16,    F16,   F16,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,     GemmPadded,        1,   256,    256,   128,    40,    64,    32,   4,   4,    2,   32,   32,     2,     4,     2,     S<2,128, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,     false,      S<2,128, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               4,               4,      false,     S<16, 16, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8,   Masking>,
         DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle<     Row,      Col,      Row,     Row,   F16,    F16,    F16,   F16,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,     GemmPadded,        1,   256,    256,   128,    40,   128,    32,   4,   4,    2,   32,   32,     2,     4,     4,     S<2,128, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,     false,      S<2,128, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               4,               4,      false,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8,   Masking>,
         DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle<     Row,      Col,      Row,     Row,   F16,    F16,    F16,   F16,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,     GemmPadded,        1,   256,    128,   256,    40,    64,    32,   4,   4,    2,   32,   32,     1,     8,     2,     S<2,128, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              4,              4,     false,      S<2,128, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               4,               4,      false,     S<16, 16, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8,   Masking>,
diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
index 8382f069d7..498bf58fb3 100644
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
@@ -40,13 +40,11 @@ template <index_t NumDimG,
           MaskingSpecialization MaskingSpec>
 using device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         // #############################################|  NumDimG| NumDimM| NumDimN| NumDimK| NumDimO|  AData|  B0Data|  B1Data|  CData|     Acc0BiasData| Acc1BiasData| AccData| CShuffle|           A|          B0|        Acc0|          B1|           C|           GEMM|   ATensorSpec|  B0TensorSpec|  B1TensorSpec|   CTensorSpec| NumGemmK| Block| Gemm01| Gemm0| Gemm0| Gemm1| Gemm1| AK1| BK1| B1K1| MPer| NPer| Gemm0| Gemm0| Gemm1|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockLds|  B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer| MaskingSpec|  D0s Bias|
         // #############################################|         |        |        |        |        |   Type|    Type|    Type|   Type|             Type|         Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Elementwise| Elementwise| Specialization|              |              |              |              | Prefetch|  Size|   MPer|  NPer|  KPer|  NPer|  KPer|    |    |     |  XDL|  XDL|  MXdl|  NXdl|  NXdl|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|    ThreadCluster|   ThreadCluster|  SrcAccessOrder|    SrcVectorDim|       SrcScalar|       DstScalar|  AddExtraN|    ThreadCluster|   ThreadCluster|  SrcAccessOrder|    SrcVectorDim|       SrcScalar|       DstScalar|  AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            | SrcScalar|
         // #############################################|         |        |        |        |        |       |        |        |       |                 |             |        |         |   Operation|   Operation|   Operation|   Operation|   Operation|               |              |              |              |              |    Stage|      |  Block| Block| Block| Block| Block|    |    |     |     |     |   Per|   Per|   Per| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          |  Lengths_K0_N_K1|    ArrangeOrder|                |                |       PerVector|    PerVector_K1|           |  Lengths_K0_N_K1|    ArrangeOrder|                |                |       PerVector|    PerVector_K1|           |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            | PerVector|
         // #############################################|         |        |        |        |        |       |        |        |       |                 |             |        |         |            |            |            |            |            |               |              |              |              |              |         |      |       |      |      |      |      |    |    |     |     |     |  Wave|  Wave|  Wave|                |               |               |               |               |               |          |                 |                |                |                |                |                |           |                 |                |                |                |                |                |           |            |            |                             |                |            |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   BF16,    BF16,    BF16,   BF16,  ck::Tuple<BF16>,  ck::Tuple<>,     F32,     BF16, PassThrough, PassThrough,    ScaleAdd, PassThrough, PassThrough,     GemmPadded, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    128,    64,    32,   128,    32,   8,   8,    2,   32,   32,     1,     2,     4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec,        1>,
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   BF16,    BF16,    BF16,   BF16,  ck::Tuple<BF16>,  ck::Tuple<>,     F32,     BF16, PassThrough, PassThrough,    ScaleAdd, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    256,   128,    32,    64,    32,   8,   8,    2,   32,   32,     2,     4,     2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S<16, 16, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec>,
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   BF16,    BF16,    BF16,   BF16,  ck::Tuple<BF16>,  ck::Tuple<>,     F32,     BF16, PassThrough, PassThrough,    ScaleAdd, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    256,   128,    32,   128,    32,   8,   8,    2,   32,   32,     2,     4,     4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec>,
diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
index b6c14d69db..744bd6456d 100644
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
@@ -40,13 +40,11 @@ template <index_t NumDimG,
           MaskingSpecialization MaskingSpec>
 using device_batched_gemm_bias_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         // #############################################|  NumDimG| NumDimM| NumDimN| NumDimK| NumDimO| AData| B0Data| B1Data| CData|    Acc0BiasData| Acc1BiasData| AccData| CShuffle|           A|          B0|        Acc0|          B1|           C|           GEMM|   ATensorSpec|  B0TensorSpec|  B1TensorSpec|   CTensorSpec| NumGemmK| Block| Gemm01| Gemm0| Gemm0| Gemm1| Gemm1| AK1| BK1| B1K1| MPer| NPer| Gemm0| Gemm0| Gemm1|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockLds|  B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer| MaskingSpec|  D0s Bias|
         // #############################################|         |        |        |        |        |  Type|   Type|   Type|  Type|            Type|         Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Elementwise| Elementwise| Specialization|              |              |              |              | Prefetch|  Size|   MPer|  NPer|  KPer|  NPer|  KPer|    |    |     |  XDL|  XDL|  MXdl|  NXdl|  NXdl|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|    ThreadCluster|   ThreadCluster|  SrcAccessOrder|    SrcVectorDim|       SrcScalar|       DstScalar|  AddExtraN|    ThreadCluster|   ThreadCluster|  SrcAccessOrder|    SrcVectorDim|       SrcScalar|       DstScalar|  AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            | SrcScalar|
         // #############################################|         |        |        |        |        |      |       |       |      |                |             |        |         |   Operation|   Operation|   Operation|   Operation|   Operation|               |              |              |              |              |    Stage|      |  Block| Block| Block| Block| Block|    |    |     |     |     |   Per|   Per|   Per| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          |  Lengths_K0_N_K1|    ArrangeOrder|                |                |       PerVector|    PerVector_K1|           |  Lengths_K0_N_K1|    ArrangeOrder|                |                |       PerVector|    PerVector_K1|           |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            | PerVector|
         // #############################################|         |        |        |        |        |      |       |       |      |                |             |        |         |            |            |            |            |            |               |              |              |              |              |         |      |       |      |      |      |      |    |    |     |     |     |  Wave|  Wave|  Wave|                |               |               |               |               |               |          |                 |                |                |                |                |                |           |                 |                |                |                |                |                |           |            |            |                             |                |            |          |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<F16>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,    ScaleAdd, PassThrough, PassThrough,     GemmPadded, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    128,    64,    32,   128,    32,   8,   8,    2,   32,   32,     1,     2,     4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec,       1>,
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<F16>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,    ScaleAdd, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    256,   128,    32,    64,    32,   8,   8,    2,   32,   32,     2,     4,     2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S<16, 16, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec>,
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<F16>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,    ScaleAdd, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    256,   128,    32,   128,    32,   8,   8,    2,   32,   32,     2,     4,     4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec>,
diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
index 2ce5124706..b342612d1c 100644
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instance.cpp
@@ -40,13 +40,11 @@ template <index_t NumDimG,
           MaskingSpecialization MaskingSpec>
 using device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_bf16_bf16_bf16_bf16_gmk_gnk_gno_gmo_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         // #############################################|  NumDimG| NumDimM| NumDimN| NumDimK| NumDimO|  AData|  B0Data|  B1Data|  CData| Acc0BiasData| Acc1BiasData| AccData| CShuffle|           A|          B0|        Acc0|          B1|           C|           GEMM|   ATensorSpec|  B0TensorSpec|  B1TensorSpec|   CTensorSpec| NumGemmK| Block| Gemm01| Gemm0| Gemm0| Gemm1| Gemm1| AK1| BK1| B1K1| MPer| NPer| Gemm0| Gemm0| Gemm1|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockTransfer| B0BlockLds|  B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockTransfer| B1BlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer| MaskingSpec|
         // #############################################|         |        |        |        |        |   Type|    Type|    Type|   Type|         Type|         Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Elementwise| Elementwise| Specialization|              |              |              |              | Prefetch|  Size|   MPer|  NPer|  KPer|  NPer|  KPer|    |    |     |  XDL|  XDL|  MXdl|  NXdl|  NXdl|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|    ThreadCluster|   ThreadCluster|  SrcAccessOrder|    SrcVectorDim|       SrcScalar|       DstScalar|  AddExtraN|    ThreadCluster|   ThreadCluster|  SrcAccessOrder|    SrcVectorDim|       SrcScalar|       DstScalar|  AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|            |
         // #############################################|         |        |        |        |        |       |        |        |       |             |             |        |         |   Operation|   Operation|   Operation|   Operation|   Operation|               |              |              |              |              |    Stage|      |  Block| Block| Block| Block| Block|    |    |     |     |     |   Per|   Per|   Per| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          |  Lengths_K0_N_K1|    ArrangeOrder|                |                |       PerVector|    PerVector_K1|           |  Lengths_K0_N_K1|    ArrangeOrder|                |                |       PerVector|    PerVector_K1|           |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|            |
         // #############################################|         |        |        |        |        |       |        |        |       |             |             |        |         |            |            |            |            |            |               |              |              |              |              |         |      |       |      |      |      |      |    |    |     |     |     |  Wave|  Wave|  Wave|                |               |               |               |               |               |          |                 |                |                |                |                |                |           |                 |                |                |                |                |                |           |            |            |                             |                |            |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   BF16,    BF16,    BF16,   BF16,  ck::Tuple<>,  ck::Tuple<>,     F32,     BF16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,     GemmPadded, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    128,    64,    32,   128,    32,   8,   8,    2,   32,   32,     1,     2,     4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec>,
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   BF16,    BF16,    BF16,   BF16,  ck::Tuple<>,  ck::Tuple<>,     F32,     BF16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    256,   128,    32,    64,    32,   8,   8,    2,   32,   32,     2,     4,     2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S<16, 16, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec>,
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   BF16,    BF16,    BF16,   BF16,  ck::Tuple<>,  ck::Tuple<>,     F32,     BF16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    256,   128,    32,   128,    32,   8,   8,    2,   32,   32,     2,     4,     4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec>,
diff --git a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
index 4e8adceb1c..6d64a2e2d6 100644
--- a/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instance.cpp
@@ -8,6 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -58,15 +59,26 @@ using device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    128,   128,    32,   128,    32,   8,   8,    2,   32,   32,     1,     4,     4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec>,
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,     64,   256,    32,   128,    32,   8,   8,    2,   16,   16,     1,    16,     8,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           8,               S<1, 16, 1,16>,               8, MaskingSpec>,
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,     64,   256,    32,    64,    32,   8,   8,    2,   16,   16,     1,    16,     4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<4, 64, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S<16, 16, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           4,               S<1, 32, 1, 8>,               8, MaskingSpec>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,     64,   256,    64,   128,    32,   8,   8,    2,   16,   16,     1,    16,     8,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<8, 32, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           8,               S<1, 16, 1,16>,               8, MaskingSpec>,
-        DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,     64,   256,    64,    64,    32,   8,   8,    2,   16,   16,     1,    16,     4,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<8, 32, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S<16, 16, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           4,               S<1, 32, 1, 8>,               8, MaskingSpec>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         // Padded fallback kernel
         DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,     GemmPadded, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,    128,   128,    64,   128,    32,   8,   8,    2,   32,   32,     1,     4,     4,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,     false,      S<8, 32, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,      false,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           2,               S<1, 32, 1, 8>,               8, MaskingSpec>
         // clang-format on
         >;
 
+// instances not working on gfx950
+template <index_t NumDimG,
+          index_t NumDimM,
+          index_t NumDimN,
+          index_t NumDimK,
+          index_t NumDimO,
+          MaskingSpecialization MaskingSpec>
+using device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instances_part2 =
+    std::tuple<
+        // clang-format off
+        DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,     64,   256,    64,   128,    32,   8,   8,    2,   16,   16,     1,    16,     8,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<8, 32, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S< 8, 32, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           8,               S<1, 16, 1,16>,               8, MaskingSpec>,
+        DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle<  NumDimG, NumDimM, NumDimN, NumDimK, NumDimO,   F16,    F16,    F16,   F16,  ck::Tuple<>,  ck::Tuple<>,     F32,      F16, PassThrough, PassThrough,       Scale, PassThrough, PassThrough,    GemmDefault, TensorDefault, TensorDefault, TensorDefault, TensorDefault,        1,   256,     64,   256,    64,    64,    32,   8,   8,    2,   16,   16,     1,    16,     4,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,      S<8, 32, 1>,      S<1, 0, 2>,      S<1, 0, 2>,               2,               8,               8,       true,     S<16, 16, 1>,      S<0, 2, 1>,      S<0, 2, 1>,               1,               4,               2,      false,           1,           4,               S<1, 32, 1, 8>,               8, MaskingSpec>
+        // clang-format on
+        >;
+
 void add_device_batched_gemm_masking_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instances(
     std::vector<std::unique_ptr<
         DeviceBatchedGemmSoftmaxGemmPermute<2,
@@ -97,6 +109,19 @@ void add_device_batched_gemm_masking_softmax_gemm_permute_xdl_cshuffle_f16_f16_f
             1,
             1,
             MaskingSpecialization::MaskOutUpperTriangle>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instances_part2<
+                2,
+                1,
+                1,
+                1,
+                1,
+                MaskingSpecialization::MaskOutUpperTriangle>{});
+    }
 }
 
 void add_device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instances(
@@ -129,6 +154,19 @@ void add_device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_g
             1,
             1,
             MaskingSpecialization::MaskDisabled>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_batched_gemm_softmax_gemm_permute_xdl_cshuffle_f16_f16_f16_f16_gmk_gnk_gno_gmo_instances_part2<
+                2,
+                1,
+                1,
+                1,
+                1,
+                MaskingSpecialization::MaskDisabled>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/conv2d_fwd/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instance.cpp b/library/src/tensor_operation_instance/gpu/conv2d_fwd/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instance.cpp
index c0c74ff7fb..86c17aacf0 100644
--- a/library/src/tensor_operation_instance/gpu/conv2d_fwd/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/conv2d_fwd/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instance.cpp
@@ -8,6 +8,7 @@
 #include "ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -40,14 +41,11 @@ static constexpr auto ConvFwdOddC =
 
 // arbitrary conv
 using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##########################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##########################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##########################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##########################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwdDefault,   256,   128,   128,     4,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwdDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwdDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwdDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -63,17 +61,24 @@ using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instances = std::tuple<
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwdDefault,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
-
-// 1x1, pad 0
-using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_p0_f16_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instances_2x = std::tuple<
+    // clang-format off
+        //##########################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##########################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##########################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##########################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwdDefault,   256,   128,   128,     4,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
+
+// 1x1, pad 0
+using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_p0_f16_instances = std::tuple<
+    // clang-format off
         //##########################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##########################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##########################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##########################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,   ConvFwd1x1P0,   256,   128,   128,     4,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,   ConvFwd1x1P0,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,   ConvFwd1x1P0,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,   ConvFwd1x1P0,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -89,17 +94,24 @@ using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_p0_f16_instances = std:
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,   ConvFwd1x1P0,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
-
-// 1x1, stride 1, pad 0
-using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_p0_f16_instances_2x = std::tuple<
+    // clang-format off
+        //##########################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##########################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##########################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##########################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,   ConvFwd1x1P0,   256,   128,   128,     4,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
+
+// 1x1, stride 1, pad 0
+using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances = std::tuple<
+    // clang-format off
         //##########################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##########################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##########################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##########################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwd1x1S1P0,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwd1x1S1P0,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwd1x1S1P0,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwd1x1S1P0,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -115,16 +127,23 @@ using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances = s
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwd1x1S1P0,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
-
-using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_odd_c_f16_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances_2x = std::tuple<
+    // clang-format off
+        //##########################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##########################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##########################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##########################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough, ConvFwd1x1S1P0,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
+
+using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_odd_c_f16_instances = std::tuple<
+    // clang-format off
         //##########################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##########################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##########################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##########################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,    ConvFwdOddC,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,    ConvFwdOddC,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,    ConvFwdOddC,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,    ConvFwdOddC,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4,  4, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  4, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -145,6 +164,16 @@ using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_odd_c_f16_instances = std::
         DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,    ConvFwdOddC,   128,    64,    64,     2,  4,   32,   32,    1,    2,     S<2, 16, 4>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<2, 16, 4>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_odd_c_f16_instances_2x = std::tuple<
+    // clang-format off
+        //##########################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##########################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##########################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##########################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,    ConvFwdOddC,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
 
 void add_device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instances(
     std::vector<std::unique_ptr<
@@ -159,6 +188,18 @@ void add_device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instances(
         instances, device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances{});
     add_device_operation_instances(
         instances, device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_odd_c_f16_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instances_2x{});
+        add_device_operation_instances(
+            instances, device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_p0_f16_instances_2x{});
+        add_device_operation_instances(
+            instances, device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances_2x{});
+        add_device_operation_instances(
+            instances, device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_odd_c_f16_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/conv2d_fwd_bias_relu/device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instance.cpp b/library/src/tensor_operation_instance/gpu/conv2d_fwd_bias_relu/device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instance.cpp
index 715dffb7ff..3402653e84 100644
--- a/library/src/tensor_operation_instance/gpu/conv2d_fwd_bias_relu/device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/conv2d_fwd_bias_relu/device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instance.cpp
@@ -8,6 +8,7 @@
 #include "ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_bias_activation_nhwc_kyxc_nhwk.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -39,14 +40,11 @@ static constexpr auto ConvFwdOddC =
 
 // arbitrary conv
 using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##########################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|           Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##########################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|  GlobalMemory| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##########################################################################################|       |        |        |        |   Operation|   Operation|   Operation| DataOperation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##########################################################################################|       |        |        |        |            |            |            |              |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwdDefault,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwdDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwdDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwdDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -62,17 +60,24 @@ using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instances = s
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwdDefault,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
-
-// 1x1, pad 0
-using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_p0_f16_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instances_2x = std::tuple<
+    // clang-format off
+        //##########################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|           Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##########################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|  GlobalMemory| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##########################################################################################|       |        |        |        |   Operation|   Operation|   Operation| DataOperation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##########################################################################################|       |        |        |        |            |            |            |              |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwdDefault,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
+
+// 1x1, pad 0
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_p0_f16_instances = std::tuple<
+    // clang-format off
         //##########################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|           Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##########################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|  GlobalMemory| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##########################################################################################|       |        |        |        |   Operation|   Operation|   Operation| DataOperation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##########################################################################################|       |        |        |        |            |            |            |              |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,   ConvFwd1x1P0,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,   ConvFwd1x1P0,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,   ConvFwd1x1P0,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,   ConvFwd1x1P0,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -88,17 +93,24 @@ using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_p0_f16_instan
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,   ConvFwd1x1P0,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
-
-// 1x1, stride 1, pad 0
-using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_p0_f16_instances_2x = std::tuple<
+    // clang-format off
+        //##########################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|           Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##########################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|  GlobalMemory| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##########################################################################################|       |        |        |        |   Operation|   Operation|   Operation| DataOperation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##########################################################################################|       |        |        |        |            |            |            |              |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,   ConvFwd1x1P0,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
+
+// 1x1, stride 1, pad 0
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances = std::tuple<
+    // clang-format off
         //##########################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|           Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##########################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|  GlobalMemory| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##########################################################################################|       |        |        |        |   Operation|   Operation|   Operation| DataOperation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##########################################################################################|       |        |        |        |            |            |            |              |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwd1x1S1P0,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwd1x1S1P0,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwd1x1S1P0,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwd1x1S1P0,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -114,17 +126,25 @@ using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_s1_p0_f16_ins
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwd1x1S1P0,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
-
-// Odd C
-using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_odd_c_f16_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##########################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|           Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##########################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|  GlobalMemory| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##########################################################################################|       |        |        |        |   Operation|   Operation|   Operation| DataOperation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##########################################################################################|       |        |        |        |            |            |            |              |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet, ConvFwd1x1S1P0,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+        // clang-format on
+        >;
+
+// Odd C
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_odd_c_f16_instances = std::tuple<
+    // clang-format off
         //##########################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|           Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##########################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|  GlobalMemory| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##########################################################################################|       |        |        |        |   Operation|   Operation|   Operation| DataOperation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##########################################################################################|       |        |        |        |            |            |            |              |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,    ConvFwdOddC,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,    ConvFwdOddC,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,    ConvFwdOddC,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,    ConvFwdOddC,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4,  4, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  4, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -145,6 +165,16 @@ using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_odd_c_f16_instanc
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,    ConvFwdOddC,   128,    64,    64,     2,  4,   32,   32,    1,    2,     S<2, 16, 4>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<2, 16, 4>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_odd_c_f16_instances_2x = std::tuple<
+    // clang-format off
+        //##########################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|           Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##########################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|  GlobalMemory| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##########################################################################################|       |        |        |        |   Operation|   Operation|   Operation| DataOperation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##########################################################################################|       |        |        |        |            |            |            |              |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,     AddRelu,     MemorySet,    ConvFwdOddC,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
 
 void add_device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instances(
     std::vector<DeviceConvFwdBiasActivationPtr<PassThrough, PassThrough, AddRelu>>& instances)
@@ -158,6 +188,21 @@ void add_device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instances(
         device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances{});
     add_device_operation_instances(
         instances, device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_odd_c_f16_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_p0_f16_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_odd_c_f16_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/conv2d_fwd_bias_relu_add/device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instance.cpp b/library/src/tensor_operation_instance/gpu/conv2d_fwd_bias_relu_add/device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instance.cpp
index 5676d77986..faac2813ba 100644
--- a/library/src/tensor_operation_instance/gpu/conv2d_fwd_bias_relu_add/device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/conv2d_fwd_bias_relu_add/device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instance.cpp
@@ -8,6 +8,7 @@
 #include "ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_bias_activation_add_nhwc_kyxc_nhwk.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -37,14 +38,11 @@ static constexpr auto ConvFwdOddC =
 
 // arbitrary conv
 using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##############################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##############################################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##############################################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwdDefault,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwdDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwdDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwdDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -60,17 +58,24 @@ using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instances
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwdDefault,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
-
-// 1x1, pad 0
-using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_1x1_p0_f16_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instances_2x = std::tuple<
+    // clang-format off
+        //##############################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##############################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##############################################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##############################################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwdDefault,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
+
+// 1x1, pad 0
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_1x1_p0_f16_instances = std::tuple<
+    // clang-format off
         //##############################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##############################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##############################################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##############################################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,   ConvFwd1x1P0,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,   ConvFwd1x1P0,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,   ConvFwd1x1P0,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,   ConvFwd1x1P0,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -86,17 +91,24 @@ using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_1x1_p0_f16_in
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,   ConvFwd1x1P0,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
-
-// 1x1, stride 1, pad 0
-using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_1x1_p0_f16_instances_2x = std::tuple<
+    // clang-format off
+        //##############################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##############################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##############################################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##############################################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,   ConvFwd1x1P0,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
+
+// 1x1, stride 1, pad 0
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances = std::tuple<
+    // clang-format off
         //##############################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##############################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##############################################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##############################################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwd1x1S1P0,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwd1x1S1P0,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwd1x1S1P0,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwd1x1S1P0,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -112,17 +124,25 @@ using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_1x1_s1_p0_f16
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwd1x1S1P0,    64,    32,    64,     4,  8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
-
-// Odd C
-using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_odd_c_f16_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##############################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##############################################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##############################################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd, ConvFwd1x1S1P0,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+        // clang-format on
+        >;
+
+// Odd C
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_odd_c_f16_instances = std::tuple<
+    // clang-format off
         //##############################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //##############################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //##############################################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //##############################################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,    ConvFwdOddC,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,    ConvFwdOddC,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,    ConvFwdOddC,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>,
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,    ConvFwdOddC,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4,  4, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  4, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 16, 1, 1, 8>,               8>,
@@ -143,6 +163,16 @@ using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_odd_c_f16_ins
         DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,    ConvFwdOddC,   128,    64,    64,     2,  4,   32,   32,    1,    2,     S<2, 16, 4>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<2, 16, 4>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 16, 1, 1, 4>,               8>
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_odd_c_f16_instances_2x = std::tuple<
+    // clang-format off
+        //##############################################################################################| InData| WeiData| OutData| AccData|          In|         Wei|         Out|    ConvForward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
+        //##############################################################################################|   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise| Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
+        //##############################################################################################|       |        |        |        |   Operation|   Operation|   Operation|               |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
+        //##############################################################################################|       |        |        |        |            |            |            |               |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
+        DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K<    F16,     F16,     F16,     F32, PassThrough, PassThrough,  AddReluAdd,    ConvFwdOddC,   256,   128,   128,     4, 16,   32,   32,    2,    2,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              1,      true,     S<4,  8, 8>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              1,      true,           1,           1,             S<1, 1, 32, 1, 1, 8>,               8>
+    // clang-format on
+    >;
 
 void add_device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instances(
     std::vector<DeviceConvFwdBiasActivationAddPtr<PassThrough, PassThrough, AddReluAdd>>& instances)
@@ -158,6 +188,22 @@ void add_device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instan
     add_device_operation_instances(
         instances,
         device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_odd_c_f16_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_1x1_p0_f16_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_odd_c_f16_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp
index 0d143b95ee..7fcab3ea46 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp
@@ -8,6 +8,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -29,15 +30,12 @@ static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecializa
 
 // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
 using device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                     Version|
         //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        2,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        2,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        2,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        2,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -87,6 +85,17 @@ using device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances = std::tu
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances_2x = std::tuple<
+    // clang-format off
+        //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                     Version|
+        //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        2,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 void add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances(
     std::vector<std::unique_ptr<
@@ -95,6 +104,12 @@ void add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp
index ce58469697..5bf0e8b7eb 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp
@@ -11,6 +11,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -55,11 +56,6 @@ using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances = std::tuple<
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    64,    32,    32,   8,   8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,    64,    32,    64,    32,   8,   8,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV2<   Row,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,        2,   256,   256,   256,    32,   8,   8,   32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        ,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,    64,   128,  32,  32,   16,   16,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              4,  LoopScheduler::Default,        PipelineVersion::v1>
-#endif
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
         ,
@@ -96,6 +92,18 @@ using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances = std::tuple<
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
+using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances_2x = std::tuple<
+    // clang-format off
+        //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,  BF16,  BF16,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,    64,    64,   128,  32,  32,   16,   16,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              4,  LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances(
     std::vector<std::unique_ptr<
@@ -110,6 +118,20 @@ void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances(
 
     add_device_operation_instances(
         instances, device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances<MNKPadding>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances_2x<GemmDefault>{});
+
+        add_device_operation_instances(
+            instances, device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances_2x<MNPadding>{});
+
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances_2x<MNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
index 3ebd0c5351..27240122de 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
@@ -10,6 +10,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -31,15 +32,12 @@ static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecializa
 
 // Compilation parameters for a[k, m] * b[k, n] = c[m, n]
 using device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffle<     Col,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffle<     Col,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   2,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle<     Col,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle<     Col,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   2,   2,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
@@ -98,6 +96,17 @@ using device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances = std::tuple<
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances_2x = std::tuple<
+    // clang-format off
+        //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemm_Xdl_CShuffle<     Col,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(
     std::vector<std::unique_ptr<
@@ -106,6 +115,12 @@ void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(
 {
     add_device_operation_instances(instances,
                                    device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
index bd0f7346f7..dd1abb6079 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
@@ -10,6 +10,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -31,15 +32,12 @@ static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecializa
 
 // Compilation parameters for a[k, m] * b[n, k] = c[m, n]
 using device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffle<     Col,      Col,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffle<     Col,      Col,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   2,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle<     Col,      Col,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle<     Col,      Col,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   2,   8,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
@@ -98,6 +96,17 @@ using device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances = std::tuple<
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances_2x = std::tuple<
+    // clang-format off
+        //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemm_Xdl_CShuffle<     Col,      Col,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(
     std::vector<std::unique_ptr<
@@ -106,6 +115,12 @@ void add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(
 {
     add_device_operation_instances(instances,
                                    device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
index a6ac333ea3..538e99b7e1 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
@@ -11,6 +11,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -35,27 +36,31 @@ static constexpr auto MNPadding = ck::tensor_operation::device::GemmSpecializati
 static constexpr auto MNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
 
 using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_generic_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,     MNKPadding,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 8>,              1,  LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,     MNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 8>,              1,  LoopScheduler::Default,        PipelineVersion::v1>
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_generic_instances_2x = std::tuple<
+    // clang-format off
+        //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,     MNKPadding,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 8>,              1,  LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 // Compilation parameters for a[m, k] * b[k, n] = c[m, n]
 template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
 using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // pipeline v1, 1 wave
         DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle<     Row,      Row,    Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
@@ -133,6 +138,12 @@ void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(
 
     add_device_operation_instances(
         instances, device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances<MNKPadding>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_generic_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp
index 852b053527..278e56f556 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp
@@ -11,6 +11,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -34,30 +35,35 @@ static constexpr auto MNKPadding  = ck::tensor_operation::device::GemmSpecializa
 
 // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
 using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_generic_instances = std::tuple<
-// clang-format off
+    // clang-format off
+        //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,     MNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 8>,              1,  LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
+// double rate mfma instances on gfx950
+using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_generic_instances_2x = std::tuple<
+    // clang-format off
         //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
-#if defined(CK_USE_AMD_MFMA_GFX950)
         //DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
         //DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,    64,   128,  32,  32,   16,   16,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              4,  LoopScheduler::Default,        PipelineVersion::v1>
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,     MNKPadding,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 8>,              1,  LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,     MNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 8>,              1,  LoopScheduler::Default,        PipelineVersion::v1>
+        DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,     MNKPadding,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 8>,              1,  LoopScheduler::Default,        PipelineVersion::v1>
     // clang-format on
     >;
 
 template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
 // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
 using device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //#####################|        |        |        |  Type|  Type|  Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //#####################|        |        |        |      |      |      |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // pipeline v1, 1 wave
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row,   F16,   F16,   F16,     F32,      F16, PassThrough, PassThrough, PassThrough,       GemmSpec,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8,  LoopScheduler::Default,        PipelineVersion::v1>,
@@ -125,6 +131,12 @@ void add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(
 
     add_device_operation_instances(
         instances, device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances<MNKPadding>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances, device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_generic_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp
index 41f6ec2bf7..caf17d55cb 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp
@@ -8,6 +8,8 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
+
 #ifdef CK_ENABLE_INT8
 namespace ck {
 namespace tensor_operation {
@@ -47,12 +49,19 @@ using device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances =
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row, int8_t, int8_t, int8_t, int32_t,  int32_t, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   128,    32,   128,    64,  16,  16,   32,   32,    1,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 4>,             16>,
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row, int8_t, int8_t, int8_t, int32_t,  int32_t, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    64,    32,    64,  16,  16,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,             16>,
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row, int8_t, int8_t, int8_t, int32_t,  int32_t, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,    64,    32,    64,    64,  16,  16,   32,   32,    1,    2,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 2>,             16>
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        ,
+        // clang-format on
+        >;
+// double rate mfma instances on gfx950
+using device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances_2x =
+    std::tuple<
+        // clang-format off
+        //#####################| ALayout| BLayout| CLayout|  AData|  BData|  CData| AccData| CShuffle|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //#####################|        |        |        |   Type|   Type|   Type|    Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //#####################|        |        |        |       |       |       |        |         |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //#####################|        |        |        |       |       |       |        |         |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row, int8_t, int8_t, int8_t, int32_t,  int32_t, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,   128,   128,   128,  32,  32,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,             16,             16,         1,           1,           1,               S<1, 64, 1, 4>,              16,  LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle<     Row,     Col,     Row, int8_t, int8_t, int8_t, int32_t,  int32_t, PassThrough, PassThrough, PassThrough,    GemmDefault,        1,   256,    64,    64,   256,  64,  64,   16,   16,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,             16,             16,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,             16,             16,         1,           1,           1,               S<1, 32, 1, 8>,               4,  LoopScheduler::Default,        PipelineVersion::v1>
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
-       // clang-format on
+        // clang-format on
         >;
 
 void add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances(
@@ -62,6 +71,12 @@ void add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances(
 {
     add_device_operation_instances(instances,
                                    device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(instances,
+                                       device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_kn_mn_interwave_pipeline_v1_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_kn_mn_interwave_pipeline_v1_instance.cpp
index 74cf837500..a64424e8ac 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_kn_mn_interwave_pipeline_v1_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_kn_mn_interwave_pipeline_v1_instance.cpp
@@ -9,7 +9,8 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for a[k, m] * b[k, n] = c[m, n]
-using Instances = std::tuple<
+using Instances =
+    std::tuple<
 // clang-format off
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -17,8 +18,6 @@ using Instances = std::tuple<
         //##########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|            |                       |         |
         //##########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|            |                       |         |
         //##########|      |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |            |                       |         |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
@@ -28,8 +27,8 @@ using Instances = std::tuple<
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>
 #endif
-    // clang-format on
-    >;
+        // clang-format on
+        >;
 
 void add_device_gemm_xdl_f16_f16_f16_km_kn_mn_interwave_pipeline_v1_instances(
     OwnerList<InstanceNT>& instances)
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_kn_mn_irregular_interwave_pipeline_v1_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_kn_mn_irregular_interwave_pipeline_v1_instance.cpp
index f2b28f3b40..0a0406baec 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_kn_mn_irregular_interwave_pipeline_v1_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_kn_mn_irregular_interwave_pipeline_v1_instance.cpp
@@ -17,8 +17,6 @@ using Instances = std::tuple<
         //###########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|            |                         |                           |
         //###########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|            |                         |                           |
         //###########|      |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |            |                         |                           |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmXdl<   F16,   F16,   F16,     F32,     Col,     Row,     Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   64,     16,    16,     4,  8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1,           1, LoopScheduler::Interwave,        PipelineVersion::v1>
 #endif
     // clang-format on
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_nk_mn_interwave_pipeline_v1_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_nk_mn_interwave_pipeline_v1_instance.cpp
index da5fefe5da..3671bea7a3 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_nk_mn_interwave_pipeline_v1_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_nk_mn_interwave_pipeline_v1_instance.cpp
@@ -9,7 +9,8 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for a[k, m] * b[n, k] = c[m, n]
-using Instances = std::tuple<
+using Instances =
+    std::tuple<
 // clang-format off
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -17,8 +18,6 @@ using Instances = std::tuple<
         //##########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|            |                       |                             |
         //##########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|            |                       |                             |
         //##########|      |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |            |                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
@@ -28,8 +27,8 @@ using Instances = std::tuple<
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,    64,     4,  8,   32,   32,    2,    1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,    64,   128,     4,  8,   32,   32,    1,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>
 #endif
-    // clang-format on
-    >;
+        // clang-format on
+        >;
 
 void add_device_gemm_xdl_f16_f16_f16_km_nk_mn_interwave_pipeline_v1_instances(
     OwnerList<InstanceNN>& instances)
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_nk_mn_irregular_interwave_pipeline_v1_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_nk_mn_irregular_interwave_pipeline_v1_instance.cpp
index b6c03b3367..95fc8ecb46 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_nk_mn_irregular_interwave_pipeline_v1_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/km_nk_mn_irregular_interwave_pipeline_v1_instance.cpp
@@ -17,8 +17,6 @@ using Instances = std::tuple<
         //###########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|            |                       |                             |
         //###########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|            |                       |                             |
         //###########|      |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |            |                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Col,      Col,    Row, PassThrough, PassThrough, PassThrough,   GemmMNPadding,   64,    16,    16,     4,  8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>
 #endif
     // clang-format on
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_kn_mn_interwave_pipeline_v1_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_kn_mn_interwave_pipeline_v1_instance.cpp
index bf271cc3c3..fa53a3bf0f 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_kn_mn_interwave_pipeline_v1_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_kn_mn_interwave_pipeline_v1_instance.cpp
@@ -9,7 +9,8 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for a[m, k] * b[k, n] = c[m, n]
-using Instances = std::tuple<
+using Instances =
+    std::tuple<
 // clang-format off
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -17,8 +18,6 @@ using Instances = std::tuple<
         //##########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|            |                       |                             |
         //##########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|            |                       |                             |
         //##########|      |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |            |                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
@@ -37,8 +36,8 @@ using Instances = std::tuple<
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,    16,    32,     4,  8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,   GemmDefault,    64,    16,    16,     4,  8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>
 #endif
-    // clang-format on
-    >;
+        // clang-format on
+        >;
 
 void add_device_gemm_xdl_f16_f16_f16_mk_kn_mn_interwave_pipeline_v1_instances(
     OwnerList<InstanceTT>& instances)
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_kn_mn_irregular_interwave_pipeline_v1_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_kn_mn_irregular_interwave_pipeline_v1_instance.cpp
index 0df59933c2..c9d1913aec 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_kn_mn_irregular_interwave_pipeline_v1_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_kn_mn_irregular_interwave_pipeline_v1_instance.cpp
@@ -17,8 +17,6 @@ using Instances = std::tuple<
         //###########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|            |                       |                             |
         //###########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|            |                       |                             |
         //###########|      |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |            |                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmXdl<   F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,    64,    16,    16,     4,  8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>
 #endif
     // clang-format on
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_nk_mn_interwave_pipeline_v1_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_nk_mn_interwave_pipeline_v1_instance.cpp
index d9260d85ab..0410eabb70 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_nk_mn_interwave_pipeline_v1_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_nk_mn_interwave_pipeline_v1_instance.cpp
@@ -17,8 +17,6 @@ using Instances = std::tuple<
         //###########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|            |                       |                             |
         //###########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|            |                       |                             |
         //###########|      |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |            |                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmXdl<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   256,   128,     4,  8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   256,   128,   256,     4,  8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,   GemmDefault,   128,   128,   128,     4,  8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
diff --git a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_interwave_pipeline_v1_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_interwave_pipeline_v1_instance.cpp
index 8b98133ada..a41919aab7 100644
--- a/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_interwave_pipeline_v1_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_f16_f16_f16/mk_nk_mn_irregular_interwave_pipeline_v1_instance.cpp
@@ -17,8 +17,6 @@ using Instances = std::tuple<
         //###########|  Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| SrcDstVectorDim|       DstScalar|            |                       |                             |
         //###########|      |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |                |       PerVector|            |                       |                             |
         //###########|      |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |                |                |            |                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmXdl<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,    64,    64,    32,     4,  8,   32,   32,    2,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   128,   144,     8,  8,   16,   16,    2,    9,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<8,  8, 4>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              2,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>,
         DeviceGemmXdl<   F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough, GemmMNPadding,   256,   128,   144,     4,  8,   16,   16,    2,    9,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,      true,     S<4, 16, 4>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              2,              2,      true,               7,               1,           1,  LoopScheduler::Interwave,        PipelineVersion::v1>
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_instance.cpp
index 9ea79b1467..33ee841354 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_instance.cpp
@@ -10,6 +10,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -38,29 +39,36 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[k, m], b[k, n], d0[m, n], d1[m, n]
 using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_generic_instance =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
         // clang-format on
         >;
-using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_instances =
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_generic_instance_2x =
     std::tuple<
-// clang-format off
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_instances =
+    std::tuple<
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   2,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   256,    32,   2,   2,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -119,19 +127,28 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn
 #endif
         // clang-format on
         >;
-
-// irregular tile size
-using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_irregular_tile_instances =
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_instances_2x =
     std::tuple<
-// clang-format off
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
+// irregular tile size
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_irregular_tile_instances =
+    std::tuple<
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -145,6 +162,18 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn
 #endif
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Col,
@@ -168,6 +197,19 @@ void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn
     add_device_operation_instances(
         instances,
         device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_kn_mn_mn_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_instance.cpp
index f6a959a447..260f9288b2 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_instance.cpp
@@ -10,6 +10,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -38,29 +39,36 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[k, m], b[n, k], d0[m, n], d1[m, n]
 using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_generic_instance =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
         // clang-format on
         >;
-using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_instances =
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_generic_instance_2x =
     std::tuple<
-// clang-format off
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_instances =
+    std::tuple<
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   2,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   256,    32,   2,   8,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -119,19 +127,28 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn
 #endif
         // clang-format on
         >;
-
-// irregular tile size
-using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_irregular_tile_instances =
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_instances_2x =
     std::tuple<
-// clang-format off
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
+// irregular tile size
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_irregular_tile_instances =
+    std::tuple<
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -145,6 +162,18 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn
 #endif
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Col,
@@ -168,6 +197,19 @@ void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn
     add_device_operation_instances(
         instances,
         device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_km_nk_mn_mn_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instance.cpp
index 8a68a8bb76..8113c7013a 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instance.cpp
@@ -10,6 +10,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -38,29 +39,36 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[m, k], b[k, n], d0[m, n], d1[m, n]
 using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_generic_instance =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    64,  16,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
         // clang-format on
         >;
-using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instances =
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_generic_instance_2x =
     std::tuple<
-// clang-format off
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    64,  16,   2,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instances =
+    std::tuple<
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -119,19 +127,28 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn
 #endif
         // clang-format on
         >;
-
-// irregular tile size
-using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_irregular_tile_instances =
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instances_2x =
     std::tuple<
-// clang-format off
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
+// irregular tile size
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_irregular_tile_instances =
+    std::tuple<
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -145,6 +162,18 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn
 #endif
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
@@ -168,6 +197,19 @@ void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn
     add_device_operation_instances(
         instances,
         device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_instance.cpp
index 42f7a10b9f..b94e902335 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_add_fastgelu/device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_instance.cpp
@@ -10,6 +10,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -38,29 +39,36 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[m, k], b[n, k], d0[m, n], d1[m ,n]
 using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_generic_instance =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
         // clang-format on
         >;
-using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_instances =
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_generic_instance_2x =
     std::tuple<
-// clang-format off
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,   128,   128,   128,    64,  16,  16,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_instances =
+    std::tuple<
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -110,19 +118,28 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn
 #endif
         // clang-format on
         >;
-
-// irregular tile size
-using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_irregular_tile_instances =
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_instances_2x =
     std::tuple<
-// clang-format off
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
+// irregular tile size
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_irregular_tile_instances =
+    std::tuple<
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -136,6 +153,18 @@ using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn
 #endif
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,   F16, PassThrough, PassThrough, AddAddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
@@ -159,6 +188,19 @@ void add_device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn
     add_device_operation_instances(
         instances,
         device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_f16_mk_nk_mn_mn_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_instance.cpp
index 6d1c16776a..b34f5e162c 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_instance.cpp
@@ -6,6 +6,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp"
 #include "ck/utility/sequence.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -23,28 +24,35 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[k, m], b[k, n], d0[m, n]
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_generic_instance =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_generic_instance_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|        Ds|      E| AData| BData| AccData| CShuffle|    DsData| EData|           A|           B|         CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|    Layout| Layout|  Type|  Type|    Type| DataType|      Type|  Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |          |       |      |      |        |         |          |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |          |       |      |      |        |         |          |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   2,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   256,    32,   2,   2,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -103,19 +111,27 @@ using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_instanc
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_instances_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|        Ds|      E| AData| BData| AccData| CShuffle|    DsData| EData|           A|           B|         CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|    Layout| Layout|  Type|  Type|    Type| DataType|      Type|  Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |          |       |      |      |        |         |          |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |          |       |      |      |        |         |          |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 // irregular tile size
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_irregular_tile_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -129,6 +145,18 @@ using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_irregul
 #endif
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Col,
@@ -151,6 +179,19 @@ void add_device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_inst
     add_device_operation_instances(
         instances,
         device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_kn_mn_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_instance.cpp
index 8b746781ef..befa868e28 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_instance.cpp
@@ -6,6 +6,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp"
 #include "ck/utility/sequence.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -23,28 +24,35 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[k, m], b[n, k], d0[m, n], d1[m, n]
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_generic_instance =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_generic_instance_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|        Ds|      E| AData| BData| AccData| CShuffle|    DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|    Layout| Layout|  Type|  Type|    Type| DataType|      Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |          |       |      |      |        |         |          |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |          |       |      |      |        |         |          |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   2,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   256,    32,   2,   8,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -103,19 +111,27 @@ using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_instanc
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_instances_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|        Ds|      E| AData| BData| AccData| CShuffle|    DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|    Layout| Layout|  Type|  Type|    Type| DataType|      Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |          |       |      |      |        |         |          |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |          |       |      |      |        |         |          |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 // irregular tile size
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_irregular_tile_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -129,6 +145,18 @@ using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_irregul
 #endif
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Col,
@@ -151,6 +179,19 @@ void add_device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_inst
     add_device_operation_instances(
         instances,
         device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_km_nk_mn_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instance.cpp
index 414ab674d3..47b3d16fcd 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instance.cpp
@@ -6,6 +6,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp"
 #include "ck/utility/sequence.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -23,28 +24,35 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[m, k], b[k, n], d0[m, n], d1[m, n]
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_generic_instance =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_generic_instance_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|        Ds|      E| AData| BData| AccData| CShuffle|    DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|    Layout| Layout|  Type|  Type|    Type| DataType|      Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |          |       |      |      |        |         |          |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |          |       |      |      |        |         |          |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -103,19 +111,27 @@ using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instanc
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instances_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|        Ds|      E| AData| BData| AccData| CShuffle|    DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|    Layout| Layout|  Type|  Type|    Type| DataType|      Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |          |       |      |      |        |         |          |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |          |       |      |      |        |         |          |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 // irregular tile size
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_irregular_tile_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -129,6 +145,18 @@ using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_irregul
 #endif
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
@@ -151,6 +179,19 @@ void add_device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_inst
     add_device_operation_instances(
         instances,
         device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_instance.cpp
index cf6a64eb26..7c9ab826ae 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_fastgelu/device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_instance.cpp
@@ -6,6 +6,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp"
 #include "ck/utility/sequence.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -23,28 +24,35 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[m, k], b[n, k], d0[m, n], d1[m ,n]
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_generic_instance =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_generic_instance_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
+
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|        Ds|      E| AData| BData| AccData| CShuffle|    DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|    Layout| Layout|  Type|  Type|    Type| DataType|      Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |          |       |      |      |        |         |          |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |          |       |      |      |        |         |          |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -94,19 +102,27 @@ using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_instanc
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_instances_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|        Ds|      E| AData| BData| AccData| CShuffle|    DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|    Layout| Layout|  Type|  Type|    Type| DataType|      Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |          |       |      |      |        |         |          |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |          |       |      |      |        |         |          |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_Tuple,   F16, PassThrough, PassThrough, AddFastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 // irregular tile size
 using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_irregular_tile_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -120,6 +136,18 @@ using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_irregul
 #endif
         // clang-format on
         >;
+// double rate mfma instances on gfx950
+using device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,     Row_Tuple,    Row,   F16,   F16,     F32,      F32,     F16_Tuple,   F16, PassThrough, PassThrough,    AddFastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
@@ -142,6 +170,19 @@ void add_device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_inst
     add_device_operation_instances(
         instances,
         device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_add_fastgelu_xdl_c_shuffle_f16_f16_f16_f16_mk_nk_mn_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_kn_mn_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_kn_mn_mn_mn_instance.cpp
index 184d7974d8..28a452c1a1 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_kn_mn_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_kn_mn_mn_mn_instance.cpp
@@ -37,13 +37,11 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[k, m], b[k, n], d0[m, n], d1[m, n], gamma[n], beta[n]
 template <LoopScheduler GemmLoopScheduler, PipelineVersion GemmPipeline>
 using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_kn_mn_mn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#######################################|      A|      B|            Ds|      H| AData| BData| AccData| CShuffle|        DsData| EMeanVarData| GammaData|  BetaData| HData|           A|           B|            CDE|           H|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|          PostShuffle|     PostShuffle|            Layernorm|       Layernorm|     LoopScheduler|     Pipeline|
         //#######################################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|         Type|      Type|      Type|  Type| Elementwise| Elementwise|    Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ThreadClusterLengths| ScalarPerVector| ThreadClusterLengths| ThreadSliceSize|                  |             |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |   Operation|   Operation|      Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|                 _M_N|   _NWaveNPerXdl|                 _M_N|              _M|                  |             |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |            |            |               |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                     |                |                     |                |                  |             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,     AddReluAdd, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   2,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,     AddReluAdd, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,     AddReluAdd, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   2,   2,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
@@ -66,13 +64,11 @@ using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_kn_mn_mn_mn_instan
 // irregular tile size
 using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_kn_mn_mn_mn_irregular_tile_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //#######################################|      A|      B|            Ds|      H| AData| BData| AccData| CShuffle|        DsData| EMeanVarData| GammaData|  BetaData| HData|           A|           B|            CDE|           H|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|           PostShuffle|     PostShuffle|            Layernorm|       Layernorm|          LoopScheduler|              Pipeline|
         //#######################################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|         Type|      Type|      Type|  Type| Elementwise| Elementwise|    Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|  ThreadClusterLengths| ScalarPerVector| ThreadClusterLengths| ThreadSliceSize|                       |                      |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |   Operation|   Operation|      Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|                  _M_N|   _NWaveNPerXdl|                 _M_N|              _M|                       |                      |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |            |            |               |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                      |                |                     |                |                       |                      |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // pipeline v1, 1 wave
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Col,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,     AddReluAdd, PassThrough, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<16, 4>,              1,             S<16, 4>,               1, LoopScheduler::Default,  PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_nk_mn_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_nk_mn_mn_mn_instance.cpp
index ef37c82c7f..13366238d6 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_nk_mn_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_nk_mn_mn_mn_instance.cpp
@@ -37,13 +37,11 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[k, m], b[k, n], d0[m, n], d1[m, n], gamma[n], beta[n]
 template <LoopScheduler GemmLoopScheduler, PipelineVersion GemmPipeline>
 using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_nk_mn_mn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#######################################|      A|      B|            Ds|      H| AData| BData| AccData| CShuffle|        DsData| EMeanVarData| GammaData|  BetaData| HData|           A|           B|         CDE|           H|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|          PostShuffle|     PostShuffle|            Layernorm|       Layernorm|     LoopScheduler|     Pipeline|
         //#######################################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|         Type|      Type|      Type|  Type| Elementwise| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ThreadClusterLengths| ScalarPerVector| ThreadClusterLengths| ThreadSliceSize|                  |             |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |   Operation|   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|                 _M_N|   _NWaveNPerXdl|                 _M_N|              _M|                  |             |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |            |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                     |                |                     |                |                  |             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   2,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   2,   8,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
@@ -66,13 +64,11 @@ using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_nk_mn_mn_mn_instan
 // irregular tile size
 using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_km_nk_mn_mn_mn_irregular_tile_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //#######################################|      A|      B|            Ds|      H| AData| BData| AccData| CShuffle|        DsData| EMeanVarData| GammaData|  BetaData| HData|           A|           B|         CDE|           H|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|          PostShuffle|     PostShuffle|            Layernorm|       Layernorm|          LoopScheduler|               Pipeline|
         //#######################################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|         Type|      Type|      Type|  Type| Elementwise| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ThreadClusterLengths| ScalarPerVector| ThreadClusterLengths| ThreadSliceSize|                       |                       |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |   Operation|   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|                 _M_N|   _NWaveNPerXdl|                 _M_N|              _M|                       |                       |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |            |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                     |                |                     |                |                       |                       |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // pipeline v1, 1 wave
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Col,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,              S<16, 4>,              1,             S<16, 4>,               1, LoopScheduler::Default,   PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_kn_mn_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_kn_mn_mn_mn_instance.cpp
index 40fbc85be0..8a4889ee83 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_kn_mn_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_kn_mn_mn_mn_instance.cpp
@@ -37,13 +37,11 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[k, m], b[k, n], d0[m, n], d1[m, n], gamma[n], beta[n]
 template <LoopScheduler GemmLoopScheduler, PipelineVersion GemmPipeline>
 using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_kn_mn_mn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#######################################|      A|      B|            Ds|      H| AData| BData| AccData| CShuffle|        DsData| EMeanVarData| GammaData|  BetaData| HData|           A|           B|         CDE|           H|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|          PostShuffle|     PostShuffle|            Layernorm|       Layernorm|     LoopScheduler|     Pipeline|
         //#######################################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|         Type|      Type|      Type|  Type| Elementwise| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ThreadClusterLengths| ScalarPerVector| ThreadClusterLengths| ThreadSliceSize|                  |             |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |   Operation|   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|                 _M_N|   _NWaveNPerXdl|                 _M_N|              _M|                  |             |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |            |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                     |                |                     |                |                  |             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
@@ -66,13 +64,11 @@ using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_kn_mn_mn_mn_instan
 // irregular tile size
 using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_kn_mn_mn_mn_irregular_tile_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //#######################################|      A|      B|            Ds|      H| AData| BData| AccData| CShuffle|        DsData| EMeanVarData| GammaData|  BetaData| HData|           A|           B|         CDE|           H|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|          PostShuffle|     PostShuffle|            Layernorm|       Layernorm|          LoopScheduler|              Pipeline|
         //#######################################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|         Type|      Type|      Type|  Type| Elementwise| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ThreadClusterLengths| ScalarPerVector| ThreadClusterLengths| ThreadSliceSize|                       |                      |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |   Operation|   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|                 _M_N|   _NWaveNPerXdl|                 _M_N|              _M|                       |                      |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |            |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                     |                |                     |                |                       |                      |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // pipeline v1, 1 wave
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Row,    Row, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<16, 4>,              1,             S<16, 4>,               1, LoopScheduler::Default,  PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
diff --git a/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_nk_mn_mn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_nk_mn_mn_mn_instance.cpp
index 464279c376..fc3cbcf905 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_nk_mn_mn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_add_relu_add_layernorm/device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_nk_mn_mn_mn_instance.cpp
@@ -37,13 +37,11 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // input: a[k, m], b[k, n], d0[m, n], d1[m, n], gamma[n], beta[n]
 template <LoopScheduler GemmLoopScheduler, PipelineVersion GemmPipeline>
 using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_nk_mn_mn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#######################################|      A|      B|            Ds|      H| AData| BData| AccData| CShuffle|        DsData| EMeanVarData| GammaData|  BetaData| HData|           A|           B|         CDE|           H|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|          PostShuffle|     PostShuffle|            Layernorm|       Layernorm|     LoopScheduler|     Pipeline|
         //#######################################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|         Type|      Type|      Type|  Type| Elementwise| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ThreadClusterLengths| ScalarPerVector| ThreadClusterLengths| ThreadSliceSize|                  |             |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |   Operation|   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|                 _M_N|   _NWaveNPerXdl|                 _M_N|              _M|                  |             |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |            |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                     |                |                     |                |                  |             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,             S<32, 8>,               8,             S<32, 8>,               1, GemmLoopScheduler, GemmPipeline>,
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,             S<16, 8>,               8,             S<16, 8>,               1, GemmLoopScheduler, GemmPipeline>,
@@ -63,13 +61,11 @@ using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_nk_mn_mn_mn_instan
 // irregular tile size
 using device_gemm_add_relu_add_xdl_c_shuffle_layernorm_f16_mk_nk_mn_mn_mn_irregular_tile_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //#######################################|      A|      B|            Ds|      H| AData| BData| AccData| CShuffle|        DsData| EMeanVarData| GammaData|  BetaData| HData|           A|           B|         CDE|           H|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|          PostShuffle|     PostShuffle|            Layernorm|       Layernorm|          LoopScheduler|              Pipeline|
         //#######################################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|         Type|      Type|      Type|  Type| Elementwise| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ThreadClusterLengths| ScalarPerVector| ThreadClusterLengths| ThreadSliceSize|                       |                      |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |   Operation|   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|                 _M_N|   _NWaveNPerXdl|                 _M_N|              _M|                       |                      |
         //#######################################|       |       |              |       |      |      |        |         |              |             |          |          |      |            |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                     |                |                     |                |                       |                      |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // pipeline v1, 1 wave
         DeviceGemmMultipleDLayernorm_Xdl_CShuffle<    Row,    Col, Row_Row_Tuple,    Row,   F16,   F16,     F32,      F32, F16_F16_Tuple,          F16,       F16,       F16,   F16, PassThrough, PassThrough,  AddReluAdd, PassThrough, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<16, 4>,              1,             S<16, 4>,               1, LoopScheduler::Default,  PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
diff --git a/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
index e2bf62ca94..c5e82abed6 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
@@ -6,6 +6,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp"
 #include "ck/utility/sequence.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,28 +23,34 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // outout: e[m, n]
 // input: a[k, m], b[k, n]
 using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_generic_instance = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
     // clang-format on
     >;
-using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_generic_instance_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
+
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances = std::tuple<
+    // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu,    GemmDefault,        1,   256,   256,   128,    32,   2,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu,    GemmDefault,        1,   256,   128,   256,    32,   2,   2,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -102,18 +109,26 @@ using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances = std::t
 #endif
     // clang-format on
     >;
-
-// irregular tile size
-using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_irregular_tile_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
+
+// irregular tile size
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_irregular_tile_instances = std::tuple<
+    // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -127,6 +142,18 @@ using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_irregular_tile_ins
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Col,
@@ -148,6 +175,18 @@ void add_device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances(
     add_device_operation_instances(
         instances,
         device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances, device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_kn_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
index bbf2836cf8..1e65127726 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
@@ -6,6 +6,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp"
 #include "ck/utility/sequence.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,28 +23,34 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // outout: e[m, n]
 // input: a[k, m], b[k, n]
 using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_generic_instance = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
     // clang-format on
     >;
-using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_generic_instance_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
+
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances = std::tuple<
+    // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu,    GemmDefault,        1,   256,   256,   128,    32,   2,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu,    GemmDefault,        1,   256,   128,   256,    32,   2,   8,   32,   32,    2,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              2,         0,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -102,18 +109,26 @@ using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances = std::t
 #endif
     // clang-format on
     >;
-
-// irregular tile size
-using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_irregular_tile_instances = std::tuple<
-// clang-format off
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
+
+// irregular tile size
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_irregular_tile_instances = std::tuple<
+    // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -127,6 +142,18 @@ using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_irregular_tile_ins
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Col,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Col,
@@ -148,6 +175,18 @@ void add_device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances(
     add_device_operation_instances(
         instances,
         device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances, device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_km_nk_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
index f22739e1a6..a46af7cb16 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
@@ -6,6 +6,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp"
 #include "ck/utility/sequence.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,28 +23,34 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // outout: e[m, n]
 // input: a[m, k], b[k, n]
 using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_generic_instance = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_generic_instance_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
+
 using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|         CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F32, Empty_Tuple,   F16, PassThrough, PassThrough,    FastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F32, Empty_Tuple,   F16, PassThrough, PassThrough,    FastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   2,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F32, Empty_Tuple,   F16, PassThrough, PassThrough,    FastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F32, Empty_Tuple,   F16, PassThrough, PassThrough,    FastGelu,    GemmDefault,        1,   256,   128,   256,    32,   8,   2,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,         0,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -102,18 +109,26 @@ using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances = std::t
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|         CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,   F16,     F32,      F32, Empty_Tuple,   F16, PassThrough, PassThrough,    FastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 // irregular tile size
 using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_irregular_tile_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -127,6 +142,18 @@ using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_irregular_tile_ins
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Row,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
@@ -148,6 +175,18 @@ void add_device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances(
     add_device_operation_instances(
         instances,
         device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances, device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp
index b049a5bac1..bb0ee323d8 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_fastgelu/device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp
@@ -6,6 +6,7 @@
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp"
 #include "ck/utility/sequence.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,28 +23,34 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // outout: e[m, n]
 // input: a[m, k], b[n, k]
 using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_generic_instance = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_generic_instance_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
+
 using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|         CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F32, Empty_Tuple,   F16, PassThrough, PassThrough,    FastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F32, Empty_Tuple,   F16, PassThrough, PassThrough,    FastGelu,    GemmDefault,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F32, Empty_Tuple,   F16, PassThrough, PassThrough,    FastGelu,    GemmDefault,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F32, Empty_Tuple,   F16, PassThrough, PassThrough,    FastGelu,    GemmDefault,        1,   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 16, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>,
@@ -93,18 +100,26 @@ using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances = std::t
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances_2x = std::tuple<
+    // clang-format off
+        //##############################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|         CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col, Empty_Tuple,    Row,   F16,   F16,     F32,      F32, Empty_Tuple,   F16, PassThrough, PassThrough,    FastGelu,    GemmDefault,        1,   256,   128,   128,    64,  16,  16,   32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         1,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              8,              8,         1,           1,           1,               S<1, 32, 1, 8>,               8, LoopScheduler::Default,        PipelineVersion::v1>
+    // clang-format on
+    >;
 
 // irregular tile size
 using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_irregular_tile_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
         //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
         //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
         // pipeline v1, 1 wave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    32,   8,   8,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
 #if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES        
         // pipeline v1, 2 waves
@@ -118,6 +133,18 @@ using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_irregular_tile_ins
 #endif
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+using device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_irregular_tile_instances_2x =
+    std::tuple<
+        // clang-format off
+        //##############################|      A|      B|            Ds|      E| AData| BData| AccData| CShuffle|        DsData| EData|           A|           B|            CDE|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|          LoopScheduler|                    Pipeline|
+        //##############################| Layout| Layout|        Layout| Layout|  Type|  Type|    Type| DataType|          Type|  Type| Elementwise| Elementwise|    Elementwise| Specialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |   Operation|   Operation|      Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|                       |                            |
+        //##############################|       |       |              |       |      |      |        |         |              |      |            |            |               |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |                       |                            |
+        // pipeline v1, 1 wave
+        DeviceGemmMultipleD_Xdl_CShuffle<    Row,    Col,   Empty_Tuple,    Row,   F16,   F16,     F32,      F32,   Empty_Tuple,   F16, PassThrough, PassThrough,       FastGelu, GemmMNKPadding,        1,    64,    16,    16,    64,  16,  16,   16,   16,    1,    1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              1,              8,         1,     S<4, 16, 1>,     S<1, 0, 2>,     S<1, 0, 2>,             2,              1,              8,         1,           1,           1,               S<1, 16, 1, 4>,               1, LoopScheduler::Default,        PipelineVersion::v1>
+        // clang-format on
+        >;
 
 void add_device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(
     std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
@@ -139,6 +166,18 @@ void add_device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances(
     add_device_operation_instances(
         instances,
         device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_irregular_tile_instances{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_generic_instance_2x{});
+        add_device_operation_instances(
+            instances, device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances_2x{});
+        add_device_operation_instances(
+            instances,
+            device_gemm_fastgelu_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_irregular_tile_instances_2x{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_v1_interwave_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_v1_interwave_instance.cpp
index 385f474ecc..efc7a7ebfd 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_v1_interwave_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_v1_interwave_instance.cpp
@@ -33,13 +33,11 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // Compilation parameters for a[m, k] * b[k, n] = c[m, n]
 template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
 using device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_v1_iw_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C|          GEMM| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //#########################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         //PipelineVersion::v1; interwave
         DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,  F16, PipelineVersion::v1, LoopScheduler::Interwave>,
         DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,  F16, PipelineVersion::v1, LoopScheduler::Interwave>,
diff --git a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_v1_irregular_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_v1_irregular_instance.cpp
index 0b48bbf606..6a323d323f 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_v1_irregular_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_v1_irregular_instance.cpp
@@ -34,13 +34,11 @@ template <ck::tensor_operation::device::GemmSpecialization GemmSpec,
           ck::PipelineVersion PipVer,
           ck::LoopScheduler LoopSche>
 using device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_irregular_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C|          GEMM| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //#########################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,      GemmSpec,   128,    16,   128,     4,  8,   16,   16,    1,    4,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               4,  F16, PipVer, LoopSche>,
         DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,      GemmSpec,   128,    16,   256,     4,  8,   16,   16,    1,    8,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               4,  F16, PipVer, LoopSche>,
         DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,      Row,    Row, PassThrough, PassThrough, PassThrough,      GemmSpec,   256,    16,   256,     4,  8,   16,   16,    1,    4,  S<1, 4, 16, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,      true,           1,           1,                   S<1, 16, 1, 16>,              4,  F16, PipVer, LoopSche>,        
diff --git a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_v1_interwave_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_v1_interwave_instance.cpp
index 422db05b35..2855235f97 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_v1_interwave_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_v1_interwave_instance.cpp
@@ -33,14 +33,12 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
 // Compilation parameters for a[m, k] * b[k, n] = c[m, n]
 template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
 using device_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_v1_iw_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################|AData| BData| CData| AccData| ALayout| BLayout| CLayout|           A|           B|           C|          GEMM| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|
         //#########################| Type|  Type|  Type|    Type|        |        |        | Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
         //#########################|     |      |      |        |        |        |        |   Operation|   Operation|   Operation|              |      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|
         //#########################|     |      |      |        |        |        |        |            |            |            |              |      |      |      |      |   |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |
         //PipelineVersion::v1; interwave
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,  F16, PipelineVersion::v1, LoopScheduler::Interwave>,
         DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 32, 1, 8>,               8,  F16, PipelineVersion::v1, LoopScheduler::Interwave>,
         DeviceGemmXdlSplitKCShuffle<  F16,   F16,   F16,     F32,     Row,     Col,     Row, PassThrough, PassThrough, PassThrough,      GemmSpec,   128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,      true,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             3,              8,              8,      true,           1,           1,                   S<1, 16, 1, 8>,               8,  F16, PipelineVersion::v1, LoopScheduler::Interwave>,
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_kn_mn.hpp
index 5540d2d884..59154f3439 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_kn_mn.hpp
@@ -36,13 +36,11 @@ static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
 template <GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_bf16_bf16_bf16_km_kn_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Col,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  32,   32,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,               1,              8,              4,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   4,   4,  32,   32,    2,    2,     S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,               1,              8,              4,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   2,   2,  32,   32,    2,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,               1,              4,              2,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
@@ -60,13 +58,11 @@ using device_gemm_xdl_universal_bf16_bf16_bf16_km_kn_mn_comp_instances = std::tu
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_bf16_bf16_bf16_km_kn_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly
         DeviceGemm_Xdl_CShuffleV3<  Col,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   4,   4,  16,   16,    1,    1,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,               1,              4,              4,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   2,   2,  16,   16,    1,    1,     S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,               1,              2,              2,          0,    S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              2,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp
index d9f9969621..6f0d722b60 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp
@@ -36,17 +36,12 @@ static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
 template <GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Compute friendly
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   8,  32,   32,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              8,              4,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  32,   32,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   2,   2,  32,   32,    4,    4,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              2,          0,    S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   4,   8,  32,   32,    2,    2,     S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,              1,              8,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
@@ -56,9 +51,6 @@ using device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances = std::tu
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   8,  32,   32,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              8,              4,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  32,   32,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   2,   2,  32,   32,    4,    4,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              2,          0,    S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 32, 1, 8>,                4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   4,   8,  32,   32,    2,    2,     S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,              1,              8,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   4,   8,  32,   32,    2,    2,     S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,              1,              8,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   4,   8,  32,   32,    2,    2,     S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,              1,              8,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
@@ -66,16 +58,22 @@ using device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances = std::tu
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   2,   2,  32,   32,    2,    2,     S<32, 8, 1>,    S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              2,          0,    S<32, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   8,  32,   32,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              8,              4,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
+        DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 32, 1, 8>,                4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly 
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   4,   8,  16,   16,    1,    1,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   4,   4,  16,   16,    1,    1,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,          0,    S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
@@ -90,15 +88,19 @@ using device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances = std::tup
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   4,   8,  16,   16,    1,    1,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,    64,   4,   8,  16,   16,    1,    1,     S<16, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              4,          0,    S<8,  8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,    64,   4,   8,  16,   16,    1,    1,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,    64,   4,   8,  16,   16,    1,    2,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,    64,   4,   4,  16,   16,    1,    2,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,          0,    S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,  128,    64,   4,   8,  16,   16,    1,    4,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    16,  256,    64,   2,   8,  16,   16,    1,    4,     S<32, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              2,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,              4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    16,  256,    64,   2,   2,  16,   16,    1,    4,     S<32, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              2,          0,    S<32, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,          1,           1,                   S<1, 16, 1, 16>,              4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
     // clang-format on
     >;
+// instances not working on gfx950
+template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Col,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,    64,   4,   8,  16,   16,    1,    2,     S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              4,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
+    // clang-format on
+    >;
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_default_instance.cpp
index 02272e84b3..5396c16cb8 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_default_instance
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances_part2<GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_kpadding_instance.cpp
index 89acbb6f68..5cc2d987bc 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,13 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_kpadding_instanc
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances_part2<GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_mkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_mkpadding_instance.cpp
index 52227620b4..3f7e2fe302 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_mkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_mkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_mkpadding_instan
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances<GemmMKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances_part2<
+                GemmMKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_mpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_mpadding_instance.cpp
index 311b1d0b28..550fbce684 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_mpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_mpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,13 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_mpadding_instanc
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances<GemmMPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_comp_instances_part2<GemmMPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_default_instance.cpp
index 40a9239ad8..01fcabc872 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_default_instan
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances<Intrawave, GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances_part2<Intrawave,
+                                                                                  GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_kpadding_instance.cpp
index 3a9dc7b081..22ad32465e 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_kpadding_insta
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances<Intrawave, GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances_part2<Intrawave,
+                                                                                  GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_mkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_mkpadding_instance.cpp
index 8fef199b25..335bbb6d36 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_mkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_mkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -17,6 +18,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v1_mkpadding_inst
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances<Intrawave,
                                                                         GemmMKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances_part2<Intrawave,
+                                                                                  GemmMKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_default_instance.cpp
index 4f2c95e20d..3102f7c158 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_default_instan
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances<Interwave, GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances_part2<Interwave,
+                                                                                  GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_kpadding_instance.cpp
index c58cfb592f..7c0ce3019f 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_kpadding_insta
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances<Interwave, GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances_part2<Interwave,
+                                                                                  GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_mkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_mkpadding_instance.cpp
index 7cd3bfc882..e54c611e73 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_mkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_mkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -17,6 +18,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_v2_mkpadding_inst
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances<Interwave,
                                                                         GemmMKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_km_nk_mn_mem_instances_part2<Interwave,
+                                                                                  GemmMKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp
index 0619a98cf0..50fdca9348 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp
@@ -36,39 +36,37 @@ static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
 template <GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  32,   32,    4,    4,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   2,   2,  32,   32,    4,    4,     S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              2,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,          0,          1,           2,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,          0,          1,           2,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   128,   8,   4,  16,   16,    1,    1,     S<16, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<32, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,   128,   8,   4,  16,   16,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
@@ -83,13 +81,18 @@ using device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances = std::tup
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,    64,   8,   4,  16,   16,    1,    1,     S<8,  8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,    64,   4,   4,  16,   16,    1,    1,     S<16, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<16, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,    64,   8,   4,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,    64,   8,   4,  16,   16,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,  128,    64,   8,   4,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    16,  256,    64,   8,   4,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,              4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
     // clang-format on
     >;
+// instances not working on gfx950
+template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,    64,   8,   4,  16,   16,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,  128,    64,   8,   4,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
+    // clang-format on
+    >;
+
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp
index 246e7a5067..e86d0a6ff9 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_default_instance
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp
index 9737dd5f0c..5f619d2850 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,13 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instanc
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
index 01f1315646..6a1ad1e890 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_insta
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmMNKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<
+                GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
index 0d1cb4f25b..65c1cbce96 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instan
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmMNPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<
+                GemmMNPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_default_instance.cpp
index 377e2f90a6..d98dd7b99d 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_default_instan
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances<Intrawave, GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances_part2<Intrawave,
+                                                                                  GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_kpadding_instance.cpp
index 2ad1d1e52f..9da5e8b1cd 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_kpadding_insta
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances<Intrawave, GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances_part2<Intrawave,
+                                                                                  GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp
index f82fb92302..fea562d208 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -17,6 +18,15 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v1_mnkpadding_ins
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances<Intrawave,
                                                                         GemmMNKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances_part2<
+                Intrawave,
+                GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_default_instance.cpp
index 2f8abf0a88..cc05571615 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_default_instan
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances<Interwave, GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances_part2<Interwave,
+                                                                                  GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_kpadding_instance.cpp
index 2e0f670aad..939adc4c90 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_kpadding_insta
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances<Interwave, GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances_part2<Interwave,
+                                                                                  GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp
index d7dc599748..0f3d081cf7 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -17,6 +18,15 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_v2_mnkpadding_ins
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances<Interwave,
                                                                         GemmMNKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_mem_instances_part2<
+                Interwave,
+                GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp
index 3edbd28cd8..7d141a47e1 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp
@@ -36,44 +36,42 @@ static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
 template <GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Compute friendly
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  32,   32,    4,    4,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   2,   2,  32,   32,    4,    4,     S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
-        // AGPR Spill
-        // DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
-        // AGPR Spill when use permuted lds layout. so, use padding for these two.
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 32, 1, 8>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   192,    64,   8,   8,  32,   32,    4,    3,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
+        // AGPR Spill
+        // DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
+        // AGPR Spill when use permuted lds layout. so, use padding for these two.
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 32, 1, 8>,                4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   192,    64,   8,   8,  32,   32,    4,    3,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,                4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly 
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    64,   16,   128,   8,   8,  16,   16,    1,    1,     S<16, 16, 1>,    S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 64, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,   128,   8,   8,  16,   16,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
@@ -90,13 +88,17 @@ using device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances = std::tup
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   8,   8,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,    64,   8,   8,  16,   16,    1,    1,     S<8,  8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8,  8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,    64,   8,   8,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,    64,   8,   8,  16,   16,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,  128,    64,   8,   8,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    16,  256,    64,   8,   8,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,              4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
     // clang-format on
     >;
+// instances not working on gfx950
+template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,    64,   8,   8,  16,   16,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,  128,    64,   8,   8,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
+    // clang-format on
+    >;
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp
index cbfaf9aaa0..d30ce71d6a 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instance
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances_part2<GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instance.cpp
index 3b0ccf9d83..b908cacb4b 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,13 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instanc
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances_part2<GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_default_instance.cpp
index 3800b7955d..301395c4e6 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_default_instan
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances<Intrawave, GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances_part2<Intrawave,
+                                                                                  GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_kpadding_instance.cpp
index 9e78df7b5e..ddaa504345 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v1_kpadding_insta
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances<Intrawave, GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances_part2<Intrawave,
+                                                                                  GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_default_instance.cpp
index 21f118ea60..b08137c17f 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_default_instan
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances<Interwave, GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances_part2<Interwave,
+                                                                                  GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_kpadding_instance.cpp
index 3e841a8498..ed553738b1 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_v2_kpadding_insta
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances<Interwave, GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_mem_instances_part2<Interwave,
+                                                                                  GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp
index e3f3afff46..940da94e70 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp
@@ -34,22 +34,17 @@ static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
 template <GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  32,   32,    4,    4,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   2,   2,  32,   32,    4,    4,     S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              2,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   256,    32,   8,   4,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
@@ -58,16 +53,21 @@ using device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances = std::tuple
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   2,   2,  32,   32,    2,    2,     S<32, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<32,  8, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              2,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   128,   8,   4,  16,   16,    1,    1,     S<16, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<32, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,   128,   8,   4,  16,   16,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<32, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_default_instance.cpp
index 41d6481c9f..8e5ce4ded4 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_default_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances_part2<GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_kpadding_instance.cpp
index de41821d98..1922e2b9ab 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_kpadding_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances_part2<GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
index cdde9fa43c..8cfcb85aa0 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnkpadding_instance
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances<GemmMNKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances_part2<GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnpadding_instance.cpp
index 04237cc62c..e502314388 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnpadding_instances
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances<GemmMNPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_instances_part2<GemmMNPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp
index e39c9a63b9..d83014d5e8 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp
@@ -34,13 +34,11 @@ static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
 template <GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Compute friendly
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  32,   32,    4,    4,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
@@ -57,13 +55,9 @@ using device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_instances = std::tuple
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  16,   16,    8,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   2,   2,  16,   16,    8,    8,     S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
         // AGPR Spill
         // DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         // AGPR Spill when use permuted lds layout. so, use padding for these two.
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 64, 1, 4>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   256,    32,   8,   8,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
@@ -71,16 +65,22 @@ using device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_instances = std::tuple
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 64, 1, 4>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly 
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    64,   16,   128,   8,   8,  16,   16,    1,    1,     S<16, 16, 1>,    S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 64, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,   128,   8,   8,  16,   16,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
@@ -104,10 +104,6 @@ using device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances = std::tuple<
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,    64,   8,   8,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,    64,   4,   4,  16,   16,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,    64,   2,   2,  16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<32, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,    64,   8,   8,  16,   16,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,  128,    64,   8,   8,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   64,    64,   8,   8,  32,   32,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,  128,    64,   8,   8,  32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    16,  256,    64,   8,   8,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,              4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
@@ -116,6 +112,14 @@ using device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances = std::tuple<
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    32,  256,    64,   2,   2,  32,   32,    1,    2,     S<32, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<32, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,          1,           1,                   S<1, 16, 1, 16>,              8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
     // clang-format on
     >;
+// instances not working on gfx950
+template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   64,    64,   8,   8,  16,   16,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,  128,    64,   8,   8,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
+    // clang-format on
+    >;
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_default_instance.cpp
index 77addd6ad2..ecae32aa18 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_default_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_instances_part2<GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp
index 4fb034d3b0..f65f47a08a 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_kpadding_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_comp_instances_part2<GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v1_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v1_default_instance.cpp
index 7f7ec14ba6..3d16b5c282 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v1_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v1_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v1_default_instances
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances<Intrawave, GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances_part2<Intrawave,
+                                                                               GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v1_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v1_kpadding_instance.cpp
index 32634a6129..b834022c91 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v1_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v1_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v1_kpadding_instance
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances<Intrawave, GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances_part2<Intrawave,
+                                                                               GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v2_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v2_default_instance.cpp
index 3062add942..50063c81ad 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v2_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v2_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v2_default_instances
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances<Interwave, GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances_part2<Interwave,
+                                                                               GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v2_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v2_kpadding_instance.cpp
index ede5e4c428..99469f5436 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v2_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v2_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -16,6 +17,14 @@ void add_device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_v2_kpadding_instance
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances<Interwave, GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f16_f16_mk_nk_mn_mem_instances_part2<Interwave,
+                                                                               GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn.hpp
index 43dc6be076..ff13de1d6a 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn.hpp
@@ -35,36 +35,36 @@ static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
 template <GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         // Disable due to test failure
         // DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   4,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   256,    32,   8,   4,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   4,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   256,   8,   4,  16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F16,   F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    16,   32,   256,   8,   4,  16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_default_instance.cpp
index f3e96e83f8..0a86fd3f9e 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_default_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_instances_part2<GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_kpadding_instance.cpp
index d73b75fcdc..e88928efbc 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_kpadding_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_instances_part2<GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
index 19894a4402..05b3e3762d 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_mnkpadding_instances
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_instances<GemmMNKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_instances_part2<GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_mnpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_mnpadding_instance.cpp
index f1123e5715..83db356dc8 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_mnpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_mnpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_mnpadding_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_instances<GemmMNPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f8_f16_mk_kn_mn_comp_instances_part2<GemmMNPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn.hpp
index 9bdb2f51c2..bb10da37f4 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn.hpp
@@ -35,33 +35,33 @@ static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
 template <GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Compute friendly
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,  16,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,  16,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,            16,             16,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly 
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,   128,   8,   16,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F16,    F8,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   128,   8,   16,  16,   16,    1,    1,     S<16, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 8, 1>,      S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_default_instance.cpp
index b6d916f26c..bae0d6f7c6 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_default_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_instances_part2<GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_kpadding_instance.cpp
index e72a748e96..906368f54d 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f16_f8_f16/device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_kpadding_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f16_f8_f16_mk_nk_mn_comp_instances_part2<GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn.hpp
index 616133e1ba..315a6446c1 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn.hpp
@@ -47,11 +47,6 @@ using device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_instances = std::tuple<
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,   128,  16,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<32,  8, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,             16,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4, F8>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,  16,   4,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<16, 16, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4, F8>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,   128,  16,   8,  16,   16,    8,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<16, 16, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,             16,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    64,  16,   4,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<16, 16, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,             16,              4,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,   128,  16,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<16, 16, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,             16,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,   128,  16,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<32,  8, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,             16,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,   128,  16,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<32,  8, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,             16,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5, F8>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   256,    64,  16,   4,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<16, 16, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,             16,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1, F8>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   128,    64,  16,   4,  32,   32,    4,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<16, 16, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1, F8>,
@@ -61,6 +56,23 @@ using device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_instances = std::tuple<
 #endif
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_instances_part2 = std::tuple<
+// clang-format off
+        //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
+        //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
+        //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
+        //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
+#if defined(__gfx94__) || defined(CK_USE_GFX94) || defined(CK_USE_FP8_ON_UNSUPPORTED_ARCH)
+        //Only enable these instances on gfx94x
+        // Compute friendly
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    64,  16,   4,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<16, 16, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,             16,              4,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,   128,  16,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<16, 16, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,             16,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Row,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,   128,  16,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<32,  8, 1>,    S<0, 2, 1>,    S<0, 2, 1>,               1,             16,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>
+#endif
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_mem_instances = std::tuple<
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_default_instance.cpp
index 96f171e066..e2f843ac58 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_default_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_instances_part2<GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_kpadding_instance.cpp
index 4965fe51c6..3778848db3 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_kpadding_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_instances_part2<GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_nkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_nkpadding_instance.cpp
index d325c47d8a..2e6224c139 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_nkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_nkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_nkpadding_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_instances<GemmNKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f8_f8_bf16_mk_kn_mn_comp_instances_part2<GemmNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn.hpp
index 6388c13444..27d7933477 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn.hpp
@@ -42,16 +42,7 @@ using device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_instances = std::tuple<
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
 #if defined(__gfx94__) || defined(CK_USE_GFX94) || defined(CK_USE_FP8_ON_UNSUPPORTED_ARCH)
         // Compute friendly
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    64,  16,  16,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4, F8>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,   128,  16,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4, F8>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,  16,  16,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4, F8>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,   128,  16,  16,  16,   16,    8,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    64,  16,  16,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    128, 16,  16,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    128, 16,  16,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          2,           1,                   S<1, 64, 1, 4>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
-        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   192,    128, 16,  16,  32,   32,    4,    3,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    128, 16,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    128, 16,  16,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5, F8>,
         DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   256,    64,  16,  16,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1, F8>,
@@ -63,6 +54,26 @@ using device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_instances = std::tuple<
 #endif
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_instances_part2 = std::tuple<
+// clang-format off
+        //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
+        //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
+        //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
+        //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
+#if defined(__gfx94__) || defined(CK_USE_GFX94) || defined(CK_USE_FP8_ON_UNSUPPORTED_ARCH)
+        // Compute friendly
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,  16,  16,  32,   32,    2,    2,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4, F8>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    64,  16,  16,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4, F8>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,   128,  16,  16,  16,   16,    8,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    64,  16,  16,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    128, 16,  16,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          2,           1,                   S<1, 64, 1, 4>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   192,    128, 16,  16,  32,   32,    4,    3,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>,
+        DeviceGemm_Xdl_CShuffleV3<  Row,     Col,     Row,     F8,     F8,    BF16,   F32,     BF16,  PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    128, 16,  16,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,             16,             16,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3, F8>
+#endif
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_mem_instances = std::tuple<
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_default_instance.cpp
index 48581a7344..d6c9809020 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_default_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_instances_part2<GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_kpadding_instance.cpp
index 09bc544deb..fc6ad01742 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal/device_gemm_xdl_universal_f8_f8_bf16/device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -15,6 +16,13 @@ void add_device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_kpadding_instances(
 {
     add_device_operation_instances(
         instances, device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_f8_f8_bf16_mk_nk_mn_comp_instances_part2<GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_batched/device_batched_gemm_xdl_universal_bf16_bf16_bf16/device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal_batched/device_batched_gemm_xdl_universal_bf16_bf16_bf16/device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp
index 7fb690c8b2..1cc0b8c76e 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_batched/device_batched_gemm_xdl_universal_bf16_bf16_bf16/device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_batched/device_batched_gemm_xdl_universal_bf16_bf16_bf16/device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp
@@ -50,9 +50,6 @@ using device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances =
         DeviceBatchedGemmMultiD_Xdl_CShuffle_V3<  Row,     Col, DsLayout,     Row,     BF16,   BF16, DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   2,   2,  32,   32,    4,    4,     S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,          1,           1,                   S<1, 16, 1, 16>,         S<4>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceBatchedGemmMultiD_Xdl_CShuffle_V3<  Row,     Col, DsLayout,     Row,     BF16,   BF16, DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,         S<4>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceBatchedGemmMultiD_Xdl_CShuffle_V3<  Row,     Col, DsLayout,     Row,     BF16,   BF16, DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,         S<4>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceBatchedGemmMultiD_Xdl_CShuffle_V3<  Row,     Col, DsLayout,     Row,     BF16,   BF16, DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 16, 1, 16>,         S<4>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceBatchedGemmMultiD_Xdl_CShuffle_V3<  Row,     Col, DsLayout,     Row,     BF16,   BF16, DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 32, 1, 8>,          S<4>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceBatchedGemmMultiD_Xdl_CShuffle_V3<  Row,     Col, DsLayout,     Row,     BF16,   BF16, DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   160,    64,   8,   8,  16,   16,    8,    5,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 32, 1, 8>,          S<4>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceBatchedGemmMultiD_Xdl_CShuffle_V3<  Row,     Col, DsLayout,     Row,     BF16,   BF16, DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   160,    64,   8,   8,  32,   32,    1,    5,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 64, 1, 4>,          S<8>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
@@ -62,6 +59,15 @@ using device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances =
         DeviceBatchedGemmMultiD_Xdl_CShuffle_V3<  Row,     Col, DsLayout,     Row,     BF16,   BF16, DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,         S<4>,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec,
+          typename DsLayout   = ck::Tuple<>,
+          typename DsDataType = ck::Tuple<>>
+using device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceBatchedGemmMultiD_Xdl_CShuffle_V3<  Row,     Col, DsLayout,     Row,     BF16,   BF16, DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 16, 1, 16>,         S<4>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched,
           GemmSpecialization GemmSpec,
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_batched/device_batched_gemm_xdl_universal_bf16_bf16_bf16/device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_batched/device_batched_gemm_xdl_universal_bf16_bf16_bf16/device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp
index 12aa7c3804..3c17d4cbc0 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_batched/device_batched_gemm_xdl_universal_bf16_bf16_bf16/device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_batched/device_batched_gemm_xdl_universal_bf16_bf16_bf16/device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_default_
     add_device_operation_instances(
         instances,
         device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_batched_gemm_xdl_universal_bf16_bf16_bf16_mk_nk_mn_comp_instances_part2<
+                GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp
index 68c6ce89ab..6cadb7deb8 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp
@@ -39,20 +39,15 @@ template <GemmSpecialization GemmSpec,
           typename DsLayout   = ck::Tuple<>,
           typename DsDataType = ck::Tuple<>>
 using device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout|  DsLayout| CLayout| AData| BData|      DsData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|                Block-wiseGemm|
         //#########################|        |        |          |        |  Type|  Type|        Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                      Pipeline|
         //#########################|        |        |          |        |      |      |            |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                      Verision|
         //#########################|        |        |          |        |      |      |            |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                              |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   128,   256,    32,   8,   4,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
@@ -60,19 +55,26 @@ using device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances =
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec,
+          typename DsLayout   = ck::Tuple<>,
+          typename DsDataType = ck::Tuple<>>
+using device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched,
           GemmSpecialization GemmSpec,
           typename DsLayout   = ck::Tuple<>,
           typename DsDataType = ck::Tuple<>>
 using device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout|  DsLayout| CLayout|  AData| BData|      DsData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|                Block-wiseGemm|
         //#########################|        |        |          |        |   Type|  Type|        Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                      Pipeline|
         //#########################|        |        |          |        |       |      |            |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                      Verision|
         //#########################|        |        |          |        |       |      |            |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                              |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,   BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   128,    32,    16,    64,   8,   4,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,   BF16,   BF16, DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,    64,    16,    16,   128,   8,   4,  16,   16,    1,    1,     S<16, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp
index bd5cfc5f20..4be2db2531 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_default_i
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<
+                GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp
index f5c2c95a03..93fa56c1a5 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<
+                GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
index 978a027048..cd032e2a23 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_mnkpaddin
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmMNKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<
+                GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
index c8296e9bfc..3ff8d8e172 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_bf16_bf16/device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmMNPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<
+                GemmMNPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn.hpp
index 4e1e5567d5..c753a91d0a 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn.hpp
@@ -40,15 +40,11 @@ template <GemmSpecialization GemmSpec,
           typename DsLayout   = ck::Tuple<>,
           typename DsDataType = ck::Tuple<>>
 using device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| DsLayout| CLayout|AData| BData|    DsData|  CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                          Block-wiseGemm|                Block-wiseGemm|
         //#########################|        |        |         |        | Type|  Type|      Type|   Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                                Pipeline|                      Pipeline|
         //#########################|        |        |         |        |     |      |          |       |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                               Scheduler|                      Verision|
         //#########################|        |        |         |        |     |      |          |       |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                        |                              |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        //DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   128,   128,   128,  16,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
-        DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,      GemmSpec,   128,    32,   128,   128,  16,   4,  32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v2>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
@@ -59,20 +55,30 @@ using device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances = st
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,      GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,               8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// double rate mfma instances on gfx950
+template <GemmSpecialization GemmSpec,
+          typename DsLayout   = ck::Tuple<>,
+          typename DsDataType = ck::Tuple<>>
+using device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances_2x = std::tuple<
+    // clang-format off
+        //#########################| ALayout| BLayout| DsLayout| CLayout|AData| BData|    DsData|  CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                          Block-wiseGemm|                Block-wiseGemm|
+        //#########################|        |        |         |        | Type|  Type|      Type|   Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                                Pipeline|                      Pipeline|
+        //#########################|        |        |         |        |     |      |          |       |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                               Scheduler|                      Verision|
+        //#########################|        |        |         |        |     |      |          |       |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                        |                              |
+        DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,     F32,     BF16, PassThrough, PassThrough, PassThrough,      GemmSpec,   128,    32,   128,   128,  16,   4,  32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v2>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched,
           GemmSpecialization GemmSpec,
           typename DsLayout   = ck::Tuple<>,
           typename DsDataType = ck::Tuple<>>
 using device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| DsLayout| CLayout|AData| BData|     DsData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|                Block-wiseGemm|
         //#########################|        |        |         |        | Type|  Type|       Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                      Pipeline|
         //#########################|        |        |         |        |     |      |           |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                      Verision|
         //#########################|        |        |         |        |     |      |           |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                              |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        //DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   128,    32,  128,   128,  16,   4,  32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
-#else
         // Latency friendly
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,    64,    16,   16,   256,   8,   4,  16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   128,    16,   32,   256,   8,   4,  16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
@@ -85,7 +91,6 @@ using device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_mem_instances = std
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   128,    32,  128,    64,   8,   4,  32,   32,    1,    2,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,    16,  256,    64,   8,   4,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 16>,              4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row, DsLayout,     Row, BF16,   I8,  DsDataType,  BF16,   F32,     BF16,   PassThrough, PassThrough, PassThrough,      GemmSpec,   256,    32,  256,    64,   8,   4,  32,   32,    1,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              16,              4,          0,          1,           1,                   S<1, 16, 1, 16>,              8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
     // clang-format on
     >;
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_default_instance.cpp
index 66db3ddb3d..0ad6c2c75a 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_default_ins
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances_2x<
+                GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instance.cpp
index 7d3c832ee5..b5f784b9aa 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_kpadding_in
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances_2x<
+                GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
index 2759b878a3..1c48fea463 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances<GemmMNKPadding>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances_2x<
+                GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
index 04a0229300..e500ebde87 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_bf16_i8_bf16/device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_bf16_i8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_i
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances<GemmMNPadding>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_bf16_i8_bf16_mk_kn_mn_comp_instances_2x<
+                GemmMNPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp
index 928c325ab7..d4fed3b561 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp
@@ -39,20 +39,15 @@ template <GemmSpecialization GemmSpec,
           typename DsLayout   = ck::Tuple<>,
           typename DsDataType = ck::Tuple<>>
 using device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout|  DsLayout| CLayout|AData| BData|     DsData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |          |        | Type|  Type|       Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |          |        |     |      |           |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |          |        |     |      |           |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  F16,   F16, DsDataType,   F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  F16,   F16, DsDataType,   F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  F16,   F16, DsDataType,   F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  F16,   F16, DsDataType,   F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  F16,   F16, DsDataType,   F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  F16,   F16, DsDataType,   F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  F16,   F16, DsDataType,   F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  F16,   F16, DsDataType,   F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   256,    32,   8,   4,  32,   32,    2,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
@@ -60,19 +55,26 @@ using device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances = std
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  F16,   F16, DsDataType,   F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec,
+          typename DsLayout   = ck::Tuple<>,
+          typename DsDataType = ck::Tuple<>>
+using device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,     Row,  F16,   F16, DsDataType,   F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched,
           GemmSpecialization GemmSpec,
           typename DsLayout   = ck::Tuple<>,
           typename DsDataType = ck::Tuple<>>
 using device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| DsLayout| CLayout|AData| BData|    DsData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |         |        | Type|  Type|      Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |         |        |     |      |          |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |         |        |     |      |          |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,    Row,  F16,   F16, DsDataType,  F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   8,   4,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffleV3R1<     Row,     Row,  DsLayout,    Row,  F16,   F16, DsDataType,  F16,   F32,     F16,    PassThrough, PassThrough, PassThrough,       GemmSpec,    64,    16,   16,   128,   8,   4,  16,   16,    1,    1,     S<16, 4, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 4, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,          0,          1,           1,                   S<1, 16, 1, 4>,               4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_default_instance.cpp
index 53d2f8f0c3..00d88467e8 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_default_inst
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances_part2<
+                GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_kpadding_instance.cpp
index 11f696c509..a7e60a6d40 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_kpadding_ins
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances_part2<
+                GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
index 34e129bdcb..7ce358c695 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_mnkpadding_i
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances<GemmMNKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances_part2<
+                GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnpadding_instance.cpp
index 7a580b6716..e64c145987 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_reduce/device_gemm_xdl_universal_f16_f16_f16/device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn_comp_mnpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_f16_f16_f16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -24,6 +25,14 @@ void add_device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_mnpadding_in
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances<GemmMNPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_reduce_f16_f16_f16_mk_kn_mn_comp_instances_part2<
+                GemmMNPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn.hpp
index 2a02995827..5353fe16b5 100755
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn.hpp
@@ -48,14 +48,18 @@ using device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_instances =
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,          0,          1,           2,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              8,          0,          1,           2,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_mem_instances = std::tuple<
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp
old mode 100755
new mode 100644
index 8b2bfb5d26..9d95e6ea5a
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,6 +23,14 @@ void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_default_
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<
+                GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp
old mode 100755
new mode 100644
index a7c33ffdc4..3e54a4bb5b
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,6 +23,14 @@ void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_kpadding
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<
+                GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
old mode 100755
new mode 100644
index adc2f23d40..544723ef6a
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,6 +23,14 @@ void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnkpaddi
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmMNKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<
+                GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
old mode 100755
new mode 100644
index 0336f64665..e6959c0945
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,6 +23,14 @@ void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_mnpaddin
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_instances<GemmMNPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_kn_mn_comp_instances_part2<
+                GemmMNPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn.hpp
index 425c2c0391..209d8f644e 100755
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn.hpp
@@ -51,11 +51,7 @@ using device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_instances =
         // AGPR Spill
         // DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  16,   16,    8,    8,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          1,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
         // AGPR Spill when use permuted lds layout. so, use padding for these two.
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 32, 1, 8>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 32, 1, 8>,                4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
 
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
@@ -65,6 +61,13 @@ using device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_instances =
 
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     BF16,   BF16,  BF16,   F32,     BF16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 16, 1, 16>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_mem_instances = std::tuple<
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp
old mode 100755
new mode 100644
index e192bf14c5..924ed814ce
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,6 +23,14 @@ void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_default_
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_instances_part2<
+                GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instance.cpp
old mode 100755
new mode 100644
index d58ec3eb33..daed25130f
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_bf16_bf16_bf16/device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,6 +23,14 @@ void add_device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_kpadding
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_streamk_bf16_bf16_bf16_mk_nk_mn_comp_instances_part2<
+                GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn.hpp
old mode 100755
new mode 100644
index aa193417d9..282cea7563
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn.hpp
@@ -34,14 +34,11 @@ static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
 template <GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(__gfx950__)
-        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   32,   128,  16,   2,  32,   32,    2,    1,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<32, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              2,          0,          1,           1,                   S<1, 32, 1, 8>,               4,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v2>
-#else        
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   4,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  32,   32,    4,    4,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   2,   2,  32,   32,    4,    4,     S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              2,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
@@ -57,19 +54,16 @@ using device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_comp_instances = st
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   4,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   4,   4,  32,   32,    2,    2,     S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<16, 16, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   2,   2,  32,   32,    2,    2,     S<32, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<32,  8, 1>,    S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              2,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
-#endif // defined(__gfx950__)
     // clang-format on
     >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(__gfx950__)
-#else        
         // Latency friendly
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   8,   4,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   4,   4,  16,   16,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<16, 8, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              2,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
@@ -100,7 +94,6 @@ using device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_mem_instances = std
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    16,  256,    64,   8,   4,  16,   16,    1,    4,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,              4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    16,  256,    64,   4,   4,  16,   16,    1,    4,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              4,              4,          0,          1,           1,                   S<1, 16, 1, 16>,              4,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Row,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,    32,  256,    64,   8,   4,  32,   32,    1,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,          0,          1,           1,                   S<1, 16, 1, 16>,              8,  BlkGemmPipeSched, BlockGemmPipelineVersion::v2>
-#endif // defined(__gfx950__)
     // clang-format on
     >;
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp
old mode 100755
new mode 100644
index a685c4f252..7335a9851f
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp
@@ -34,13 +34,11 @@ static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
 template <GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|                         Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|                               Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|                              Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                                       |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Compute friendly
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   8,   8,  32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<4, 64, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  32,   32,    4,    4,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
@@ -58,9 +56,6 @@ using device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances = st
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   4,   4,  16,   16,    8,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   256,    32,   2,   2,  16,   16,    8,    8,     S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,    S<16,16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              2,              2,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
      
-#if !defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
-#endif // !defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   256,   224,    64,   8,   8,  16,   16,    8,    7,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          2,           1,                   S<1, 64, 1, 4>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
@@ -69,16 +64,21 @@ using device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances = st
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   128,   128,    64,   8,   8,  32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Interwave, BlockGemmPipelineVersion::v1>
     // clang-format on
     >;
+// instances not working on gfx950
+template <GemmSpecialization GemmSpec>
+using device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances_part2 = std::tuple<
+    // clang-format off
+        DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   256,   224,   256,    64,   8,   8,  16,   16,    7,    8,     S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 32, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           2,                   S<1, 32, 1, 8>,               8,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v3>
+    // clang-format on
+    >;
 
 template <BlockGemmPipelineScheduler BlkGemmPipeSched, GemmSpecialization GemmSpec>
 using device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //#########################| ALayout| BLayout| CLayout|AData| BData| CData| AccData| Cshuffle|           A|           B|           C|          GEMM| Block|  MPer|  NPer|  KPer| AK1| BK1|MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|     CBlockTransferClusterLengths|  CBlockTransfer|    Block-wiseGemm|               Block-wiseGemm|
         //#########################|        |        |        | Type|  Type|  Type|    Type|     Type| Elementwise| Elementwise| Elementwise|Specialization|  Size| Block| Block| Block|    |    | XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|          Pipeline|                     Pipeline|
         //#########################|        |        |        |     |      |      |        |         |   Operation|   Operation|   Operation|              |      |      |      |      |    |    |    |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl|   _NWaveNPerXdl|         Scheduler|                     Verision|
         //#########################|        |        |        |     |      |      |        |         |            |            |            |              |      |      |      |      |    |    |    |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                                 |                |                  |                             |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly 
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   8,   8,  16,   16,    1,    1,     S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,    S<8, 16, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              8,              8,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         DeviceGemm_Xdl_CShuffle_Streamk_V3<  Row,     Col,     Row,     F16,   F16,  F16,   F32,     F16,      PassThrough, PassThrough, PassThrough,       GemmSpec,   128,    32,   16,    64,   4,   4,  16,   16,    1,    1,     S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,    S<16, 8, 1>,     S<1, 0, 2>,    S<1, 0, 2>,               2,              4,              4,          0,          1,           1,                   S<1, 16, 1, 8>,               2,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_default_instance.cpp
index 546f909b3c..4f048b6525 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,6 +23,14 @@ void add_device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_default_ins
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances<GemmDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances_part2<
+                GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp
index d91de96be3..c547487983 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,6 +23,14 @@ void add_device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_kpadding_in
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances<GemmKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances_part2<
+                GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnkpadding_instance.cpp
index c70678b449..28afafa26d 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,6 +23,14 @@ void add_device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnkpadding_
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances<GemmMNKPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances_part2<
+                GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnpadding_instance.cpp
index 5410a0cc25..e9bbc4d732 100644
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -22,6 +23,14 @@ void add_device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnpadding_i
     add_device_operation_instances(
         instances,
         device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances<GemmMNPadding>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_instances_part2<
+                GemmMNPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instance.cpp
index 65e233ce08..a5bf286ca7 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instance.cpp
@@ -3,6 +3,7 @@
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -31,6 +32,30 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_bf16_comp_instances(
                                                         Empty_Tuple,
                                                         NGKHW,
                                                         ConvFwdDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_part2<2,
+                                                                  NGCHW,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NGKHW,
+                                                                  ConvFwdDefault>{});
+    }
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_2x<2,
+                                                               NGCHW,
+                                                               GKYXC,
+                                                               Empty_Tuple,
+                                                               NGKHW,
+                                                               ConvFwdDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instance.cpp
index 36c8e3cb13..ea6ba831f1 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instance.cpp
@@ -3,6 +3,7 @@
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -31,6 +32,30 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_f16_comp_instances(
                                                        Empty_Tuple,
                                                        NGKHW,
                                                        ConvFwdDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_part2<2,
+                                                                 NGCHW,
+                                                                 GKYXC,
+                                                                 Empty_Tuple,
+                                                                 NGKHW,
+                                                                 ConvFwdDefault>{});
+    }
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_2x<2,
+                                                              NGCHW,
+                                                              GKYXC,
+                                                              Empty_Tuple,
+                                                              NGKHW,
+                                                              ConvFwdDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_comp_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_comp_instance.cpp
index d98b89c556..8f0a5ca425 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_comp_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_comp_instance.cpp
@@ -3,6 +3,7 @@
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -31,6 +32,30 @@ void add_device_grouped_conv2d_fwd_xdl_ngchw_gkyxc_ngkhw_int8_comp_instances(
                                                         Empty_Tuple,
                                                         NGKHW,
                                                         ConvFwdDefault>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_int8_comp_instances_part2<2,
+                                                                  NGCHW,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NGKHW,
+                                                                  ConvFwdDefault>{});
+    }
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_int8_comp_instances_2x<2,
+                                                               NGCHW,
+                                                               GKYXC,
+                                                               Empty_Tuple,
+                                                               NGKHW,
+                                                               ConvFwdDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
index 9f06347350..a344e35c8d 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_bf16_comp_instance.cpp
@@ -3,6 +3,7 @@
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -56,6 +57,84 @@ void add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_bf16_comp_instances(
                                                                                    Empty_Tuple,
                                                                                    NHWGK,
                                                                                    ConvFwdOddC>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_part2<2,
+                                                                  NHWGC,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGK,
+                                                                  ConvFwdDefault>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_part2<2,
+                                                                  NHWGC,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGK,
+                                                                  ConvFwd1x1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_part2<2,
+                                                                  NHWGC,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGK,
+                                                                  ConvFwd1x1S1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_part2<2,
+                                                                  NHWGC,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGK,
+                                                                  ConvFwdOddC>{});
+    }
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_2x<2,
+                                                               NHWGC,
+                                                               GKYXC,
+                                                               Empty_Tuple,
+                                                               NHWGK,
+                                                               ConvFwdDefault>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_2x<2,
+                                                               NHWGC,
+                                                               GKYXC,
+                                                               Empty_Tuple,
+                                                               NHWGK,
+                                                               ConvFwd1x1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_2x<2,
+                                                               NHWGC,
+                                                               GKYXC,
+                                                               Empty_Tuple,
+                                                               NHWGK,
+                                                               ConvFwd1x1S1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_2x<2,
+                                                               NHWGC,
+                                                               GKYXC,
+                                                               Empty_Tuple,
+                                                               NHWGK,
+                                                               ConvFwdOddC>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f16_comp_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f16_comp_instance.cpp
index 9b1c7ef65e..30a8b60bfc 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f16_comp_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f16_comp_instance.cpp
@@ -3,6 +3,7 @@
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -56,6 +57,84 @@ void add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_f16_comp_instances(
                                                                                   Empty_Tuple,
                                                                                   NHWGK,
                                                                                   ConvFwdOddC>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_part2<2,
+                                                                 NHWGC,
+                                                                 GKYXC,
+                                                                 Empty_Tuple,
+                                                                 NHWGK,
+                                                                 ConvFwdDefault>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_part2<2,
+                                                                 NHWGC,
+                                                                 GKYXC,
+                                                                 Empty_Tuple,
+                                                                 NHWGK,
+                                                                 ConvFwd1x1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_part2<2,
+                                                                 NHWGC,
+                                                                 GKYXC,
+                                                                 Empty_Tuple,
+                                                                 NHWGK,
+                                                                 ConvFwd1x1S1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_part2<2,
+                                                                 NHWGC,
+                                                                 GKYXC,
+                                                                 Empty_Tuple,
+                                                                 NHWGK,
+                                                                 ConvFwdOddC>{});
+    }
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_2x<2,
+                                                              NHWGC,
+                                                              GKYXC,
+                                                              Empty_Tuple,
+                                                              NHWGK,
+                                                              ConvFwdDefault>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_2x<2,
+                                                              NHWGC,
+                                                              GKYXC,
+                                                              Empty_Tuple,
+                                                              NHWGK,
+                                                              ConvFwd1x1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_2x<2,
+                                                              NHWGC,
+                                                              GKYXC,
+                                                              Empty_Tuple,
+                                                              NHWGK,
+                                                              ConvFwd1x1S1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_2x<2,
+                                                              NHWGC,
+                                                              GKYXC,
+                                                              Empty_Tuple,
+                                                              NHWGK,
+                                                              ConvFwdOddC>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_int8_comp_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_int8_comp_instance.cpp
index 78c2257b9a..6acbb7475c 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_int8_comp_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/comp/device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_int8_comp_instance.cpp
@@ -3,6 +3,7 @@
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -56,6 +57,84 @@ void add_device_grouped_conv2d_fwd_xdl_nhwgc_gkyxc_nhwgk_int8_comp_instances(
                                                                                    Empty_Tuple,
                                                                                    NHWGK,
                                                                                    ConvFwdOddC>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_int8_comp_instances_part2<2,
+                                                                  NHWGC,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGK,
+                                                                  ConvFwdDefault>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_int8_comp_instances_part2<2,
+                                                                  NHWGC,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGK,
+                                                                  ConvFwd1x1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_int8_comp_instances_part2<2,
+                                                                  NHWGC,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGK,
+                                                                  ConvFwd1x1S1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_int8_comp_instances_part2<2,
+                                                                  NHWGC,
+                                                                  GKYXC,
+                                                                  Empty_Tuple,
+                                                                  NHWGK,
+                                                                  ConvFwdOddC>{});
+    }
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_int8_comp_instances_2x<2,
+                                                               NHWGC,
+                                                               GKYXC,
+                                                               Empty_Tuple,
+                                                               NHWGK,
+                                                               ConvFwdDefault>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_int8_comp_instances_2x<2,
+                                                               NHWGC,
+                                                               GKYXC,
+                                                               Empty_Tuple,
+                                                               NHWGK,
+                                                               ConvFwd1x1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_int8_comp_instances_2x<2,
+                                                               NHWGC,
+                                                               GKYXC,
+                                                               Empty_Tuple,
+                                                               NHWGK,
+                                                               ConvFwd1x1S1P0>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_int8_comp_instances_2x<2,
+                                                               NHWGC,
+                                                               GKYXC,
+                                                               Empty_Tuple,
+                                                               NHWGK,
+                                                               ConvFwdOddC>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
index 6fa4bc6e46..5e78cf41d8 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
@@ -3,6 +3,7 @@
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_merged_groups_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -23,23 +24,46 @@ void add_device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_bf16_inst
                                                                 PassThrough,
                                                                 PassThrough>>>& instances)
 {
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_fwd_xdl_merged_groups_bf16_instances<2,
-                                                                 NHWGC,
-                                                                 GKYXC,
-                                                                 Empty_Tuple,
-                                                                 NHWGK,
-                                                                 ConvFwdDefault>{});
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_merged_groups_bf16_instances_2x<2,
+                                                                        NHWGC,
+                                                                        GKYXC,
+                                                                        Empty_Tuple,
+                                                                        NHWGK,
+                                                                        ConvFwdDefault>{});
 
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_fwd_xdl_merged_groups_bf16_instances<2,
-                                                                 NHWGC,
-                                                                 GKYXC,
-                                                                 Empty_Tuple,
-                                                                 NHWGK,
-                                                                 ConvFwd3x3>{});
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_merged_groups_bf16_instances_2x<2,
+                                                                        NHWGC,
+                                                                        GKYXC,
+                                                                        Empty_Tuple,
+                                                                        NHWGK,
+                                                                        ConvFwd3x3>{});
+    }
+    else
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_merged_groups_bf16_instances<2,
+                                                                     NHWGC,
+                                                                     GKYXC,
+                                                                     Empty_Tuple,
+                                                                     NHWGK,
+                                                                     ConvFwdDefault>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_merged_groups_bf16_instances<2,
+                                                                     NHWGC,
+                                                                     GKYXC,
+                                                                     Empty_Tuple,
+                                                                     NHWGK,
+                                                                     ConvFwd3x3>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_f16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_f16_instance.cpp
index 9fa56f48c7..fb494acd93 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/xdl/merged_groups/device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_f16_instance.cpp
@@ -3,6 +3,7 @@
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_merged_groups_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -23,23 +24,46 @@ void add_device_grouped_conv2d_fwd_xdl_merged_groups_nhwgc_gkyxc_nhwgk_f16_insta
                                                                 PassThrough,
                                                                 PassThrough>>>& instances)
 {
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_fwd_xdl_merged_groups_f16_instances<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                ConvFwdDefault>{});
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_merged_groups_f16_instances_2x<2,
+                                                                       NHWGC,
+                                                                       GKYXC,
+                                                                       Empty_Tuple,
+                                                                       NHWGK,
+                                                                       ConvFwdDefault>{});
 
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_fwd_xdl_merged_groups_f16_instances<2,
-                                                                NHWGC,
-                                                                GKYXC,
-                                                                Empty_Tuple,
-                                                                NHWGK,
-                                                                ConvFwd3x3>{});
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_merged_groups_f16_instances_2x<2,
+                                                                       NHWGC,
+                                                                       GKYXC,
+                                                                       Empty_Tuple,
+                                                                       NHWGK,
+                                                                       ConvFwd3x3>{});
+    }
+    else
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_merged_groups_f16_instances<2,
+                                                                    NHWGC,
+                                                                    GKYXC,
+                                                                    Empty_Tuple,
+                                                                    NHWGK,
+                                                                    ConvFwdDefault>{});
+
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_merged_groups_f16_instances<2,
+                                                                    NHWGC,
+                                                                    GKYXC,
+                                                                    Empty_Tuple,
+                                                                    NHWGK,
+                                                                    ConvFwd3x3>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/comp/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/comp/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instance.cpp
index efc4640603..a94f687ef8 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/comp/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/comp/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instance.cpp
@@ -3,6 +3,7 @@
 
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -46,6 +47,62 @@ void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_comp_instances(
                                                         Empty_Tuple,
                                                         NDHWGK,
                                                         ConvFwd1x1S1P0>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_part2<3,
+                                                                  NDHWGC,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGK,
+                                                                  ConvFwdDefault>{});
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_part2<3,
+                                                                  NDHWGC,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGK,
+                                                                  ConvFwd1x1P0>{});
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_part2<3,
+                                                                  NDHWGC,
+                                                                  GKZYXC,
+                                                                  Empty_Tuple,
+                                                                  NDHWGK,
+                                                                  ConvFwd1x1S1P0>{});
+    }
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_2x<3,
+                                                               NDHWGC,
+                                                               GKZYXC,
+                                                               Empty_Tuple,
+                                                               NDHWGK,
+                                                               ConvFwdDefault>{});
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_2x<3,
+                                                               NDHWGC,
+                                                               GKZYXC,
+                                                               Empty_Tuple,
+                                                               NDHWGK,
+                                                               ConvFwd1x1P0>{});
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_bf16_comp_instances_2x<3,
+                                                               NDHWGC,
+                                                               GKZYXC,
+                                                               Empty_Tuple,
+                                                               NDHWGK,
+                                                               ConvFwd1x1S1P0>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/comp/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/comp/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_instance.cpp
index 3f3cd4b7d2..0c63345e7f 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/comp/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd/xdl/comp/device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_instance.cpp
@@ -3,6 +3,7 @@
 
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_fwd/device_grouped_conv_fwd_xdl_comp_instance.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -46,6 +47,62 @@ void add_device_grouped_conv3d_fwd_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_instances(
                                                        Empty_Tuple,
                                                        NDHWGK,
                                                        ConvFwd1x1S1P0>{});
+
+    if(ck::get_device_name() != "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_part2<3,
+                                                                 NDHWGC,
+                                                                 GKZYXC,
+                                                                 Empty_Tuple,
+                                                                 NDHWGK,
+                                                                 ConvFwdDefault>{});
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_part2<3,
+                                                                 NDHWGC,
+                                                                 GKZYXC,
+                                                                 Empty_Tuple,
+                                                                 NDHWGK,
+                                                                 ConvFwd1x1P0>{});
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_part2<3,
+                                                                 NDHWGC,
+                                                                 GKZYXC,
+                                                                 Empty_Tuple,
+                                                                 NDHWGK,
+                                                                 ConvFwd1x1S1P0>{});
+    }
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_2x<3,
+                                                              NDHWGC,
+                                                              GKZYXC,
+                                                              Empty_Tuple,
+                                                              NDHWGK,
+                                                              ConvFwdDefault>{});
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_2x<3,
+                                                              NDHWGC,
+                                                              GKZYXC,
+                                                              Empty_Tuple,
+                                                              NDHWGK,
+                                                              ConvFwd1x1P0>{});
+        add_device_operation_instances(
+            instances,
+            device_grouped_conv_fwd_xdl_f16_comp_instances_2x<3,
+                                                              NDHWGC,
+                                                              GKZYXC,
+                                                              Empty_Tuple,
+                                                              NDHWGK,
+                                                              ConvFwd1x1S1P0>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_splitk_f16_f8_f16_mk_kn_mn_irregular_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_splitk_f16_f8_f16_mk_kn_mn_irregular_instance.cpp
index 0bd53706be..20f19bd774 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_splitk_f16_f8_f16_mk_kn_mn_irregular_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_gemm/device_grouped_gemm_xdl_splitk_f16_f8_f16_mk_kn_mn_irregular_instance.cpp
@@ -30,13 +30,11 @@ using Empty_Tuple = ck::Tuple<>;
 using PassThrough                    = ck::tensor_operation::element_wise::PassThrough;
 static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
 using device_grouped_gemm_xdl_splitk_f16_f8_f16_mk_kn_mn_irregular_tile_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|           C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
         //################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise| Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|   Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
         //################################|       |       |            |       |      |      |        |         |            |      |            |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,    F8,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
         DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,    F8,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 64, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              4,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
         DeviceGroupedGemmXdlSplitKCShuffle<    Row,    Row, Empty_Tuple,    Row,   F16,    F8,     F32,      F16, Empty_Tuple,   F16, PassThrough, PassThrough, PassThrough, GemmMNKPadding,        1,   256,   192,    64,    32,   8,   8,   32,   32,    3,    1,  S<1, 4, 64, 1>,  S<0, 2, 1, 3>,  S<0, 2, 1, 3>,              3,              8,              8,         1,  S<1, 4, 32, 1>,  S<0, 1, 3, 2>,  S<0, 1, 3, 2>,             2,              2,              8,         1,           1,           1,               S<1, 32, 1, 8>,              8, PipelineVersion::v1>,
diff --git a/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp b/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp
index baf04cf12e..839d3559f7 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp
@@ -39,19 +39,31 @@ static constexpr auto GemmMNKPadding = GemmSpecialization::MNKPadding;
 static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
 static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
 
+// double rate mfma instances on gfx950
+template <typename DsLayout,
+          typename DsDataType,
+          typename CDEElementwiseOp,
+          GemmSpecialization GemmSpec = GemmMNKPadding>
+using device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances_2x = std::tuple<
+    // clang-format off
+        //###########################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|                C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
+        //###########################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise|      Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
+        //###########################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
+        //###########################################|       |       |            |       |      |      |        |         |            |      |            |            |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |   S<C,D0...,D_N| 
+        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   128,    128,  16,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,8>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v1>
+    // clang-format on
+    >;
+
 template <typename DsLayout,
           typename DsDataType,
           typename CDEElementwiseOp,
           GemmSpecialization GemmSpec = GemmMNKPadding>
 using device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances = std::tuple<
-// clang-format off
+    // clang-format off
         //###########################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|                C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
         //###########################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise|      Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //###########################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
         //###########################################|       |       |            |       |      |      |        |         |            |      |            |            |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |   S<C,D0...,D_N| 
-#if defined(CK_USE_AMD_MFMA_GFX950)
-        DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   128,    128,  16,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,8>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v1>,
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   256,   256,    32,   8,   4,   32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   128,   128,    64,   8,   4,   32,   32,    2,    2,     S<8, 32, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,    S<16, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v4>,
         // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   256,   256,   256,    32,   8,   4,   32,   32,    4,    4,     S<4, 64, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              8,              4,         0,           1,           1,               S<1, 32, 1, 8>,        S<8,8,1>,  BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion::v5>,
@@ -70,13 +82,11 @@ template <typename DsLayout,
           BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave>
 using device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_mem_instances =
     std::tuple<
-// clang-format off
+        // clang-format off
         //###########################################|      A|      B|          Ds|      E| AData| BData| AccData| CShuffle|      DsData| EData|           A|           B|                C|           GEMM| NumGemmK| Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl|  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle| CBlockTransferClusterLengths|  CBlockTransfer|
         //###########################################| Layout| Layout|      Layout| Layout|  Type|  Type|    Type| DataType|        Type|  Type| Elementwise| Elementwise|      Elementwise| Spacialization| Prefetch|  Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per|   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|         _MBlock_MWaveMPerXdl| ScalarPerVector|
         //###########################################|       |       |            |       |      |      |        |         |            |      |   Operation|   Operation|        Operation|               |    Stage|      |      |      |      |    |    |     |     | Wave| Wave| Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle|         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl|
         //###########################################|       |       |            |       |      |      |        |         |            |      |            |            |                 |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |   S<C,D0...,D_N| 
-#if defined(CK_USE_AMD_MFMA_GFX950)
-#endif // defined(CK_USE_AMD_MFMA_GFX950)
         // Latency friendly
         // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,    64,    16,    16,   256,   8,   4,   16,   16,    1,    1,     S<32, 2, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<64, 1, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 4>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
         // DeviceGroupedGemmMultipleDXdlCShuffleTileLoop<   Row,     Row,    DsLayout,    Row,  BF16,    I8,     F32,      F32,  DsDataType,  BF16, PassThrough, PassThrough, CDEElementwiseOp,       GemmSpec,        1,   128,    16,    32,   256,   8,   4,   16,   16,    1,    1,     S<32, 4, 1>,     S<1, 0, 2>,     S<1, 0, 2>,              2,              8,              8,         0,     S<64, 2, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,             16,              4,          0,          1,           1,               S<1, 16, 1, 8>,        S<4,4,1>,  BlkGemmPipeSched, BlockGemmPipelineVersion::v1>,
diff --git a/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_default_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_default_instance.cpp
index 6848774431..220cb8031f 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_default_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -27,6 +28,17 @@ void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_d
                                                                                ck::Tuple<BF16>,
                                                                                Multiply,
                                                                                GemmDefault>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances_2x<
+                ck::Tuple<Row>,
+                ck::Tuple<BF16>,
+                Multiply,
+                GemmDefault>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instance.cpp
index bb2ea76aa4..6604078013 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_kpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -27,6 +28,17 @@ void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_k
                                                                                ck::Tuple<BF16>,
                                                                                Multiply,
                                                                                GemmKPadding>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances_2x<
+                ck::Tuple<Row>,
+                ck::Tuple<BF16>,
+                Multiply,
+                GemmKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
index 7439433f8a..157b45d713 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnkpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -27,6 +28,17 @@ void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_m
                                                                                ck::Tuple<BF16>,
                                                                                Multiply,
                                                                                GemmMNKPadding>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances_2x<
+                ck::Tuple<Row>,
+                ck::Tuple<BF16>,
+                Multiply,
+                GemmMNKPadding>{});
+    }
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
index b3afed0fd7..ca7e774ccd 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_gemm_tile_loop/device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_mnpadding_instance.cpp
@@ -2,6 +2,7 @@
 // Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn.hpp"
+#include "ck/host_utility/device_prop.hpp"
 
 namespace ck {
 namespace tensor_operation {
@@ -27,6 +28,17 @@ void add_device_grouped_gemm_xdl_tile_loop_multiply_bf16_i8_bf16_mk_kn_mn_comp_m
                                                                                ck::Tuple<BF16>,
                                                                                Multiply,
                                                                                GemmMNPadding>{});
+
+    if(ck::get_device_name() == "gfx950")
+    {
+        add_device_operation_instances(
+            instances,
+            device_grouped_gemm_xdl_tile_loop_bf16_i8_bf16_mk_kn_mn_comp_instances_2x<
+                ck::Tuple<Row>,
+                ck::Tuple<BF16>,
+                Multiply,
+                GemmMNPadding>{});
+    }
 }
 
 } // namespace instance