diff --git a/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_v1.hpp b/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_v1.hpp
index 7a531f5b83..baae1c68a2 100644
--- a/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_v1.hpp
+++ b/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_v1.hpp
@@ -182,9 +182,9 @@ struct BlockwiseGemmXdlops_pipeline_bpreshuffle_v1<BlockGemmPipelineScheduler::I
 
     __device__ static constexpr auto HotLoopScheduler()
     {
-        constexpr auto num_ds_read_inst_a     = HotLoopInstList::A_LDS_Read_Inst_Num;
+        constexpr auto num_ds_read_inst_a     = HotLoopInstList::A_LDS_Read_Inst_Num * 2;
         constexpr auto num_buffer_load_inst_a = HotLoopInstList::A_Buffer_Load_Inst_Num;
-        constexpr auto num_buffer_load_inst_b = HotLoopInstList::B_Buffer_Load_Inst_Num;
+        constexpr auto num_buffer_load_inst_b = HotLoopInstList::B_Buffer_Load_Inst_Num * 2;
 
         // B global
         static_for<0, num_buffer_load_inst_b, 1>{}([&](auto i) {