how to get utilization rates of gpu? (nvml)

Question

I need gpu information for my cuda project test.

I am using nvml library, and I successfully get temperature information. But, nvml reports ERROR_NOT_SUPPORTED in nvmlDeviceGetUtilizationRates().

So now, how to get utilization rates of gpu? Clearly, there will be a way like NVIDIA GeForce Experience.

thanks,

p.s. oops! I am insufficient reputation... If you want to see NVIDIA GeForce Experience example image, click this link.

Answer 1

Maybe your graphics card, the GTX Titan (based on your image) simply does not support volatile GPU utilization?
I think this functionality is only supported on Quadro and Tesla cards.

Additionally, can you show how you use nvml in your code?

Answer 2

To see your GPU utilization in linux for nvidia cards use:

nvidia-smi

use -l to refresh for eg

nvidia-smi -l 1

this will refresh for every 1 second.