GPU usage info (nvidia-smi is not there)

Dear All!

I have used a tool python tool gpu-utils

to control how much each process running tensorflow-gpu takes memory.

It uses nvidia-smi which does not seem to be in jetson nano.

Can one install nvidia-smi on jetson nano or are there other tools to obtain info about gpu usage?

Terveisin, Markus

Hi Markus, nvidia-smi isn’t supported on Tegra-based platforms. Instead please try the tegrastats utility, you can launch it by running “sudo tegrastats” from the terminal.

See here for the documentation on tegrastats tool: https://docs.nvidia.com/jetson/l4t/index.html#page/Tegra%2520Linux%2520Driver%2520Package%2520Development%2520Guide%2FAppendixTegraStats.html%23wwconnect_header

Hi dusty_nv
I have read the documentation on tegrastats tool.
but the follow param don’t explain:

POM_5V_IN 1402/1793 POM_5V_GPU 40/45

Thanks for your help.

Hi southcamel, POM_5V_IN is the total power that the board is consuming. POM_5V_GPU is the power rail feeding the GPU, so it reports only how much power the GPU is consuming.

Hi dusty_nv

Thanks for your quickly reply.

Can we know the memory-usage on GPU, like the information shows use nvidia-smi.

Thanks very much.

try with ‘jtop’, better as sudo as themself says