diff --git a/ggml/src/ggml-cuda/softmax.cu b/ggml/src/ggml-cuda/softmax.cu
index 8f8927d250..9a78adafb5 100644
--- a/ggml/src/ggml-cuda/softmax.cu
+++ b/ggml/src/ggml-cuda/softmax.cu
@@ -407,15 +407,13 @@ static void soft_max_f32_cuda(const float *                                x,
         // Parallelize across SMs for top-p/dist-smapling
         // The heuristic for parallelizing rows across SMs vs parallelizing single row & looping over all rows was done on the basis of a B6000 GPU and
         // Can be adapted further for lower-SM-count GPUs, though keeping data in registers should be implemented first as that is the optimal solution.
-        if (ncols_x / (params.ne01 * params.ne02 * params.ne03) > 8192 && mask == nullptr && sinks == nullptr && params.scale == 1.0f && params.max_bias == 0.0f) {
-            if (ggml_cuda_info().devices[id].supports_cooperative_launch) {
+        if (ggml_cuda_info().devices[id].supports_cooperative_launch && ncols_x / (params.ne01 * params.ne02 * params.ne03) > 8192 && mask == nullptr && sinks == nullptr && params.scale == 1.0f && params.max_bias == 0.0f) {
                 ggml_cuda_pool_alloc<float> tmp_alloc(ctx.pool(), ggml_cuda_info().devices[id].nsm * sizeof(float));
 
                 void * kernel_args[] = { (void *) &x, (void *) &dst, (void *) &tmp_alloc.ptr, (void *) const_cast<soft_max_params *>(& params)};
                 CUDA_CHECK(cudaLaunchCooperativeKernel((void *) soft_max_f32_parallelize_cols,
                                                        dim3(ggml_cuda_info().devices[id].nsm, 1, 1),
                                                        dim3(WARP_SIZE * 8, 1, 1), kernel_args, 0, stream));
-            }
         } else {
             const size_t nbytes_shared_low = WARP_SIZE * sizeof(float);
             soft_max_f32<false, 0, 0>