Volatile GPU-Util is always 0

I installed CUDA with Nvidia driver.
I am using P100.

I installed tensorflow-gpu version.

I started my python neuro network application.

My CPU usage is 100%

I check GPU usage by nvidia-smi

I got 33W/250W, two processes are running and occupied 253MiB

However, ‘Volatile GPU-Util’ is always 0%

Is this a problem?

Possibly falling back to CPU, it could be that your cuda version is not supported by that tf, say cuda 9 with tf1.14. Hope helps.