So I’ve seen posts here and I’ve gotten it to work before where xorg was rendered on the iGPU and the Nvidia GPU was only powered on when I was running applications using Cuda(e.g. Tensorflow), but recently it seems to be broke? I can render xorg on my iGPU but my Nvidia GPU remains power on.
Below are the two related links.
Below is my attached log files.
nvidia-bug-report.log.gz (402.0 KB)
/sys/bus/pci/devices/0000:01:00.0/power/control
on
while it should be “auto”. This should be set by a udev rule Ubuntu installs. Please check if it’s installed /lib/udev/rules.d/50-pm-nvidia.rules
You should check documentation.
http://us.download.nvidia.com/XFree86/Linux-x86_64/460.56/README/dynamicpowermanagement.html
For some reason 50-pm-nvidia.rules is missing. I also just did a fresh reinstall of the driver and it seems on-demand is working correctly now, but if I try to do the xorg.conf file like you mentioned in the previous post I reference it doesn’t work correctly.
That user didn’t want an X on-demand setup, so following that post was pointless. Which also explains your issue: you set ‘nogpumanager’, please remove it. The gpu-manager is responsible for creating the missing udev pm file.