diff --git a/example/67_gemm_microscaling/CMakeLists.txt b/example/67_gemm_microscaling/CMakeLists.txt
index 88dca3c368..7e562f13db 100644
--- a/example/67_gemm_microscaling/CMakeLists.txt
+++ b/example/67_gemm_microscaling/CMakeLists.txt
@@ -16,7 +16,7 @@ add_example_executable(example_gemm_mx_fp4_bpreshuffle gemm_mx_fp4_bpreshuffle.c
 add_example_dependencies(example_gemm_mx example_gemm_mx_fp4_bpreshuffle)
 
 set(FP4_MXGEMM_OPTIONS)
-list(APPEND FP4_MXGEMM_OPTIONS "SHELL: -mllvm -greedy-reverse-local-assignment=1")
+list(APPEND FP4_MXGEMM_OPTIONS "SHELL: -mllvm -greedy-reverse-local-assignment=1 -mllvm --amdgpu-use-amdgpu-trackers=1")
 list(APPEND FP4_MXGEMM_OPTIONS -v --save-temps -Wno-gnu-line-marker -ftemplate-backtrace-limit=0)
 target_compile_options(example_gemm_mx_fp4 PRIVATE ${FP4_MXGEMM_OPTIONS})
 target_compile_options(example_gemm_mx_fp4_bpreshuffle PRIVATE ${FP4_MXGEMM_OPTIONS})
diff --git a/example/67_gemm_microscaling/gemm_mx_fp4_bpreshuffle.cpp b/example/67_gemm_microscaling/gemm_mx_fp4_bpreshuffle.cpp
index 960e5e23d2..5153af7b06 100644
--- a/example/67_gemm_microscaling/gemm_mx_fp4_bpreshuffle.cpp
+++ b/example/67_gemm_microscaling/gemm_mx_fp4_bpreshuffle.cpp
@@ -51,14 +51,14 @@ using DeviceOpInstance = ck::tensor_operation::device::DeviceGemmMX_Xdl_CShuffle
     ScaleBlockSize,   // ScaleBlockSize: Scaling block size
     256,               // BlockSize: Thread block size
     128,               // MPerBlock
-    256,               // NPerBlock
+    512,               // NPerBlock
     KPerBlock,        // KPerBlock
     16,               // AK1
     16,               // BK1
     16,               // MPerXDL
     16,               // NPerXDL
     8,                // MXdlPerWave
-    4,                // NXdlPerWave
+    8,                // NXdlPerWave
     S<8, 32, 1>,       // ABlockTransferThreadClusterLengths_AK0_M_AK1
     S<1, 0, 2>,       // ABlockTransferThreadClusterArrangeOrder
     S<1, 0, 2>,       // ABlockTransferSrcAccessOrder
diff --git a/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_v3_mx_bpreshuffle.hpp b/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_v3_mx_bpreshuffle.hpp
index 56a83b9f66..f3efc3e77b 100644
--- a/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_v3_mx_bpreshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_v3_mx_bpreshuffle.hpp
@@ -209,7 +209,7 @@ struct BlockwiseGemmXdlops_pipeline_v3_mx_bprehuffle<BlockGemmPipelineScheduler:
         constexpr auto num_buffer_load_total = num_buffer_load_inst_a + num_buffer_load_inst_b +
                                                num_buffer_load_a_scale + num_buffer_load_b_scale;
 
-        constexpr auto num_mfma_inst = HotLoopInstList::C_MFMA_Inst_Num;
+        constexpr auto num_mfma_inst = HotLoopInstList::C_MFMA_Inst_Num * APackedSize;
         constexpr auto mfma_cycle    = HotLoopInstList::C_MFMA_Inst_Cycle;
 
         constexpr auto ds_read_a_issue_cycle =
@@ -723,8 +723,8 @@ struct BlockwiseGemmXdlops_pipeline_v3_mx_bprehuffle<BlockGemmPipelineScheduler:
 
                     });
 
-                    // HotLoopScheduler();
-                    // __builtin_amdgcn_sched_barrier(0);
+                    HotLoopScheduler();
+                    __builtin_amdgcn_sched_barrier(0);
                 };
 
                 LoopFunc(I0, I1);