I am writing a Linux app that uses nvidia-smi and nvidia-settings to change clock speeds for linux drivers.
This works but it can be incredibly slow to apply these changes or even to read a value using nvidia-settings. It can take minutes to simply read the current clock offset for a card.
Does anyone have any idea why this is so slow?
It seems to be worse on systems with more video cards. My main machine has 12 GPUs on and that is the slowest to read or write these values.
I am on driver version 384.111