So I've got an odd one. I use iGPU for PLEX and it's being set up via direct path to /dev/dri/renderD128 in the Preferences.xml. My machine also has 2 x 3060 Ti which I use to transcode my library from h264 to h265 once in a while and play games.
The problem is the following.
I need nvidia-driver to be installed in order to use tdarr transcode, but whenever the driver is installed, PLEX (docker) ignores my iGPU and tries to transcode using nvidia, then obviously fails since no nvidia cards are being passed to it and falls back to software transcoding.
If I uninstall the nvidia driver and reboot, PLEX works as intended using the iGPU.
I am NOT binding the nvidia GPUs to VFIO at boot in order for me to play games. The way I use the server is as follows.
1. unraid sees both nvidia gpus and I use them in tdarr docker containers
2. when I start a VM with one of them passed through, the VM takes over and the GPU is not being seen anymore by the unraid
3. when I stop the VM, GPU is again being seen by unraid and can be used normally
4. iGPU is only used for PLEX
I tried to force PLEX to use the iGPU, everything is ignored. It tries to use cuda hwaccel every time no matter how I modify Preferences.xml or pass the /dev/dri to it. Some reddit posts suggested to delete codecs, which I did but still no joy.
Any ideas appreciated. Thanks
Recommended Comments
There are no comments to display.
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.