#3591 Can we improve the utilization of multiple GPUs？

Beschreibung

<img width="1248" height="840" alt="Image" src="https://github.com/user-attachments/assets/7daadf06-6abf-4cf0-9c88-66deacb2df07" />
In my dual-GPU setup, one card often hits 90% utilization, while the other is only running at around 20-30%. Is there any way to improve the workload efficiency across both GPUs? I'm not very familiar with the underlying models, so this is just a small suggestion.