Inaccurate GPU utilization display with ROCm on AMD GPUs

[MUZYOU]

Is there an existing issue for this?

• I searched the existing issues and did not find anything similar.

Current Behavior

When monitoring system resources during training sessions with ROCm on AMD GPUs, I've noticed an inconsistency in GPU utilization rates between the rocm-smi tool and the Software. While rocm-smi reports GPU utilization above 90%, the Software shows little to no change in GPU usage.

Expected Behavior

No response

Steps To Reproduce

No response

Environment

latest version through flatpak 
Intel i5 12600k and Amd 6650xt

Anything else?

No response

[MUZYOU]

[nokyan]

Hi, thanks for the issue. This is a bit odd as Resources should be incorporating the computational load of AMD GPUs. What does nvtop say?

[MUZYOU]

It shows "No GPU to monitor".

[nokyan]

That's very odd. I can go look for whatever radeontop is doing and try to adapt that maybe.