Is there a tool to monitor the real time usage of the SM or the cores inside SM

Hi,

As the title says, Is there a tool to monitor the real time usage of the SM or the cores inside SM?

I used CUDA-z, but it seems that, I cannot tell the usage of separated SMs or the separated Cores.

Thanks a lot.

Don’t think you’re going to find that level of granularity in any tool. Visual NSight or the NVIDIA profiler might provide an insight on GPU load and other similar performance metrics.