diff --git a/example/ck_tile/01_fmha/codegen/ops/fmha_batch_prefill.py b/example/ck_tile/01_fmha/codegen/ops/fmha_batch_prefill.py
index 30b9299963..76b9429b2e 100644
--- a/example/ck_tile/01_fmha/codegen/ops/fmha_batch_prefill.py
+++ b/example/ck_tile/01_fmha/codegen/ops/fmha_batch_prefill.py
@@ -470,11 +470,10 @@ def get_fwd_blobs(kernel_filter : Optional[str], receipt, optdim_list, mask_impl
                     pipelines.append(FmhaFwdPipeline('qr', 'col', 't', 't', 't', 't', logits, bias, lse, dropout, squant, mask))
                 else:
                     if bias == "bias":
-                        # TODO: rocm 6.2 compiler problem if using qr_async for bias case
-                        pipelines.append(FmhaFwdPipeline('qr', 'row', 'f', 'f', 'f', 'f', logits, bias, lse, dropout, squant, mask))
-                        pipelines.append(FmhaFwdPipeline('qr', 'row', 't', 't', 't', 't', logits, bias, lse, dropout, squant, mask))
-                        pipelines.append(FmhaFwdPipeline('qr', 'col', 'f', 'f', 'f', 'f', logits, bias, lse, dropout, squant, mask))
-                        pipelines.append(FmhaFwdPipeline('qr', 'col', 't', 't', 't', 't', logits, bias, lse, dropout, squant, mask))
+                        pipelines.append(FmhaFwdPipeline('qr_async', 'row', 'f', 'f', 'f', 'f', logits, bias, lse, dropout, squant, mask))
+                        pipelines.append(FmhaFwdPipeline('qr_async', 'row', 't', 't', 't', 't', logits, bias, lse, dropout, squant, mask))
+                        pipelines.append(FmhaFwdPipeline('qr_async', 'col', 'f', 'f', 'f', 'f', logits, bias, lse, dropout, squant, mask))
+                        pipelines.append(FmhaFwdPipeline('qr_async', 'col', 't', 't', 't', 't', logits, bias, lse, dropout, squant, mask))
                     else:
                         pipelines.append(FmhaFwdPipeline('qr_async', 'row', 't', 'f', 't', 't', logits, bias, lse, dropout, squant, mask))
                         pipelines.append(FmhaFwdPipeline('qr_async', 'row', 't', 't', 't', 't', logits, bias, lse, dropout, squant, mask))
diff --git a/include/ck_tile/ops/gemm/warp/warp_gemm.hpp b/include/ck_tile/ops/gemm/warp/warp_gemm.hpp
index f050a8e382..be5d5690ff 100644
--- a/include/ck_tile/ops/gemm/warp/warp_gemm.hpp
+++ b/include/ck_tile/ops/gemm/warp/warp_gemm.hpp
@@ -204,14 +204,6 @@ using WarpGemmMfmaBf16Bf16F32M64N4K16 = WarpGemmImpl<WarpGemmAtrributeMfmaIterat
 using WarpGemmMfma_f32_32x32x16_fp8_fp8 = WarpGemmImpl<
     WarpGemmAtrributeMfma<WarpGemmAttributeMfmaImpl_f32_32x32x16_fp8_fp8<WGAttrCtlEnum::Default_>>>;
 
-using WarpGemmMfma_f32_32x32x32_fp8_fp8 = WarpGemmImpl<WarpGemmAtrributeMfmaIterateK<
-    WarpGemmAttributeMfmaImpl_f32_32x32x16_fp8_fp8<WGAttrCtlEnum::Default_>,
-    2>>;
-
-using WarpGemmMfma_f32_32x32x32_bf8_bf8 = WarpGemmImpl<WarpGemmAtrributeMfmaIterateK<
-    WarpGemmAttributeMfmaImpl_f32_32x32x16_bf8_bf8<WGAttrCtlEnum::Default_>,
-    2>>;
-
 using WarpGemmMfma_f32_32x32x16_fp8_bf8 = WarpGemmImpl<
     WarpGemmAtrributeMfma<WarpGemmAttributeMfmaImpl_f32_32x32x16_fp8_bf8<WGAttrCtlEnum::Default_>>>;
 
@@ -221,20 +213,28 @@ using WarpGemmMfma_f32_32x32x16_bf8_fp8 = WarpGemmImpl<
 using WarpGemmMfma_f32_32x32x16_bf8_bf8 = WarpGemmImpl<
     WarpGemmAtrributeMfma<WarpGemmAttributeMfmaImpl_f32_32x32x16_bf8_bf8<WGAttrCtlEnum::Default_>>>;
 
-using WarpGemmMfma_f32_16x16x64_fp8_fp8 = WarpGemmImpl<WarpGemmAtrributeMfmaIterateK<
-    WarpGemmAttributeMfmaImpl_f32_16x16x32_fp8_fp8<WGAttrCtlEnum::Default_>,
+using WarpGemmMfma_f32_32x32x32_fp8_fp8 = WarpGemmImpl<WarpGemmAtrributeMfmaIterateK<
+    WarpGemmAttributeMfmaImpl_f32_32x32x16_fp8_fp8<WGAttrCtlEnum::Default_>,
+    2>>;
+
+using WarpGemmMfma_f32_32x32x32_bf8_bf8 = WarpGemmImpl<WarpGemmAtrributeMfmaIterateK<
+    WarpGemmAttributeMfmaImpl_f32_32x32x16_bf8_bf8<WGAttrCtlEnum::Default_>,
     2>>;
 
 using WarpGemmMfma_f32_16x16x32_fp8_fp8 = WarpGemmImpl<
     WarpGemmAtrributeMfma<WarpGemmAttributeMfmaImpl_f32_16x16x32_fp8_fp8<WGAttrCtlEnum::Default_>>>;
 
+using WarpGemmMfma_f32_16x16x32_bf8_bf8 = WarpGemmImpl<
+    WarpGemmAtrributeMfma<WarpGemmAttributeMfmaImpl_f32_16x16x32_bf8_bf8<WGAttrCtlEnum::Default_>>>;
+
+using WarpGemmMfma_f32_16x16x64_fp8_fp8 = WarpGemmImpl<WarpGemmAtrributeMfmaIterateK<
+    WarpGemmAttributeMfmaImpl_f32_16x16x32_fp8_fp8<WGAttrCtlEnum::Default_>,
+    2>>;
+
 using WarpGemmMfma_f32_16x16x64_bf8_bf8 = WarpGemmImpl<WarpGemmAtrributeMfmaIterateK<
     WarpGemmAttributeMfmaImpl_f32_16x16x32_bf8_bf8<WGAttrCtlEnum::Default_>,
     2>>;
 
-using WarpGemmMfma_f32_16x16x32_bf8_bf8 = WarpGemmImpl<
-    WarpGemmAtrributeMfma<WarpGemmAttributeMfmaImpl_f32_16x16x32_bf8_bf8<WGAttrCtlEnum::Default_>>>;
-
 using WarpGemmMfma_f32_16x16x128_fp8_fp8 = WarpGemmImpl<WarpGemmAtrributeMfma<
     WarpGemmAttributeMfmaImpl_f32_16x16x128_fp8_fp8<WGAttrCtlEnum::Default_>>>;
 
diff --git a/include/ck_tile/ops/gemm/warp/warp_gemm_attribute_mfma_impl.hpp b/include/ck_tile/ops/gemm/warp/warp_gemm_attribute_mfma_impl.hpp
index 69d22496f1..4bc4884beb 100644
--- a/include/ck_tile/ops/gemm/warp/warp_gemm_attribute_mfma_impl.hpp
+++ b/include/ck_tile/ops/gemm/warp/warp_gemm_attribute_mfma_impl.hpp
@@ -1092,7 +1092,7 @@ struct WarpGemmAttributeMfmaImpl_f32_16x16x32_f8_base
         }
         else
         {
-#if defined(__gfx94__)
+#if defined(__gfx94__) or defined(__gfx95__)
             if constexpr(std::is_same_v<ADataType, fp8_t> && std::is_same_v<BDataType, fp8_t>)
                 c_vec = __builtin_amdgcn_mfma_f32_16x16x32_fp8_fp8(
                     bit_cast<long>(a_vec), bit_cast<long>(b_vec), c_vec, 0, 0, 0);
@@ -1116,7 +1116,7 @@ struct WarpGemmAttributeMfmaImpl_f32_16x16x32_f8_base
     // c_vec = a_vec * b_vec
     CK_TILE_DEVICE CVecType operator()(const AVecType& a_vec, const BVecType& b_vec) const
     {
-#if defined(__gfx94__)
+#if defined(__gfx94__) or defined(__gfx95__)
         if constexpr(std::is_same_v<ADataType, fp8_t> && std::is_same_v<BDataType, fp8_t>)
             return bit_cast<CVecType>(__builtin_amdgcn_mfma_f32_16x16x32_fp8_fp8(
                 bit_cast<long>(a_vec), bit_cast<long>(b_vec), CVecType{0.f}, 0, 0, 0));
@@ -1251,7 +1251,7 @@ struct WarpGemmAttributeMfmaImpl_f32_32x32x16_f8_base
         }
         else
         {
-#if defined(__gfx94__)
+#if defined(__gfx94__) or defined(__gfx95__)
             if constexpr(std::is_same_v<ADataType, fp8_t> && std::is_same_v<BDataType, fp8_t>)
                 c_vec = __builtin_amdgcn_mfma_f32_32x32x16_fp8_fp8(
                     bit_cast<long>(a_vec), bit_cast<long>(b_vec), c_vec, 0, 0, 0);
@@ -1286,7 +1286,7 @@ struct WarpGemmAttributeMfmaImpl_f32_32x32x16_f8_base
     // c_vec = a_vec * b_vec
     CK_TILE_DEVICE CVecType operator()(const AVecType& a_vec, const BVecType& b_vec) const
     {
-#if defined(__gfx94__)
+#if defined(__gfx94__) or defined(__gfx95__)
         if constexpr(std::is_same_v<ADataType, fp8_t> && std::is_same_v<BDataType, fp8_t>)
             return bit_cast<CVecType>(__builtin_amdgcn_mfma_f32_32x32x16_fp8_fp8(
                 bit_cast<long>(a_vec), bit_cast<long>(b_vec), CVecType{0.f}, 0, 0, 0));