diff --git a/ggml-backend.c b/ggml-backend.c
index 06b63ae35..0764dfebc 100644
--- a/ggml-backend.c
+++ b/ggml-backend.c
@@ -1437,7 +1437,7 @@ static void sched_compute_splits(ggml_backend_sched_t sched) {
             // this is important to avoid copying constants such as KQ_mask and inp_pos multiple times
             ggml_backend_tensor_copy_async(split_backend, input, input_cpy);
         }
-        ggml_backend_synchronize(split_backend);
+        //ggml_backend_synchronize(split_backend); // necessary to measure copy time
         int64_t copy_end_us = ggml_time_us();
         copy_us[split_backend_id] += copy_end_us - copy_start_us;
 
diff --git a/ggml-vulkan.cpp b/ggml-vulkan.cpp
index c555ccc37..14fb89e09 100644
--- a/ggml-vulkan.cpp
+++ b/ggml-vulkan.cpp
@@ -4775,10 +4775,10 @@ static ggml_backend_i ggml_backend_vk_interface = {
     /* .get_name                = */ ggml_backend_vk_name,
     /* .free                    = */ ggml_backend_vk_free,
     /* .get_default_buffer_type = */ ggml_backend_vk_get_default_buffer_type,
-    /* .set_tensor_async        = */ ggml_backend_vk_set_tensor_async,
-    /* .get_tensor_async        = */ ggml_backend_vk_get_tensor_async,
-    /* .cpy_tensor_async        = */ ggml_backend_vk_cpy_tensor_async,
-    /* .synchronize             = */ ggml_backend_vk_synchronize,
+    /* .set_tensor_async        = */ NULL,  // ggml_backend_vk_set_tensor_async,
+    /* .get_tensor_async        = */ NULL,  // ggml_backend_vk_get_tensor_async,
+    /* .cpy_tensor_async        = */ NULL,  // ggml_backend_vk_cpy_tensor_async,
+    /* .synchronize             = */ NULL,  // ggml_backend_vk_synchronize,
     /* .graph_plan_create       = */ NULL,
     /* .graph_plan_free         = */ NULL,
     /* .graph_plan_compute      = */ NULL,