diff --git a/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/hsa_ext_amd.cpp b/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/hsa_ext_amd.cpp
index 5fd678784c..df93bf27d3 100644
--- a/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/hsa_ext_amd.cpp
+++ b/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/hsa_ext_amd.cpp
@@ -382,20 +382,10 @@ hsa_status_t hsa_amd_profiling_get_async_copy_time(
 
   core::Agent* agent = signal->async_copy_agent();
 
-  if (agent == NULL) {
+  if (agent == nullptr) {
     return HSA_STATUS_ERROR;
   }
 
-  // Validate the embedded agent pointer if the signal is IPC.
-  if (signal->isIPC()) {
-    for (auto it : core::Runtime::runtime_singleton_->gpu_agents()) {
-      if (it == agent) break;
-    }
-    // If the agent isn't a GPU then it is from a different process or it's a CPU.
-    // Assume it's a CPU and illegal uses will generate garbage data same as kernel completion.
-    agent = core::Runtime::runtime_singleton_->cpu_agents()[0];
-  }
-
   if (agent->device_type() == core::Agent::DeviceType::kAmdGpuDevice) {
     // Translate timestamp from GPU to system domain.
     static_cast<amd::GpuAgentInt*>(agent)->TranslateTime(signal, *time);
diff --git a/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/runtime.cpp b/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/runtime.cpp
index 488c440830..5940a51af1 100644
--- a/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/runtime.cpp
+++ b/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/runtime.cpp
@@ -471,6 +471,7 @@ hsa_status_t Runtime::CopyMemory(void* dst, core::Agent& dst_agent,
   // For cpu to cpu, fire and forget a copy thread.
   const bool profiling_enabled =
       (dst_agent.profiling_enabled() || src_agent.profiling_enabled());
+  if (profiling_enabled) completion_signal.async_copy_agent(&dst_agent);
   std::thread(
       [](void* dst, const void* src, size_t size,
          std::vector<core::Signal*> dep_signals,