diff --git a/include/cutlass/gemm/kernel/sm90_gemm_array_tma_warpspecialized_cooperative.hpp b/include/cutlass/gemm/kernel/sm90_gemm_array_tma_warpspecialized_cooperative.hpp
index 25a7e2f7..e758059e 100644
--- a/include/cutlass/gemm/kernel/sm90_gemm_array_tma_warpspecialized_cooperative.hpp
+++ b/include/cutlass/gemm/kernel/sm90_gemm_array_tma_warpspecialized_cooperative.hpp
@@ -608,10 +608,6 @@ public:
     // Get pipeline stage increments from tensor shapes
     auto k_tile_count = size<3>(gA_mkl);
 
-    // Ensure that the kernel does not touch
-    // unflushed global memory prior to this instruction
-    cutlass::arch::wait_on_dependent_grids();
-
     if (warp_group_role == WarpGroupRole::Producer) {
       cutlass::arch::warpgroup_reg_dealloc<LoadRegisterRequirement>();
 
diff --git a/include/cutlass/gemm/kernel/sm90_gemm_array_tma_warpspecialized_pingpong.hpp b/include/cutlass/gemm/kernel/sm90_gemm_array_tma_warpspecialized_pingpong.hpp
index 2c55987a..7120db78 100644
--- a/include/cutlass/gemm/kernel/sm90_gemm_array_tma_warpspecialized_pingpong.hpp
+++ b/include/cutlass/gemm/kernel/sm90_gemm_array_tma_warpspecialized_pingpong.hpp
@@ -646,10 +646,6 @@ public:
     // Get pipeline stage increments from tensor shapes
     auto k_tile_count = size<3>(gA_mkl);
 
-    // Ensure that the kernel does not touch
-    // unflushed global memory prior to this instruction
-    cutlass::arch::wait_on_dependent_grids();
-
     if (warp_group_role == WarpGroupRole::Producer) {
       cutlass::arch::warpgroup_reg_dealloc<LoadRegisterRequirement>();