I am running the latest stable release of Plex and I have an Intel HD GPU and a Nvidia P2000 video card. I am using CentOS 7.
I have the Intel HD GPU enabled and it works, however when the amount of transcode streams hits its limits, the P2000 GPU is not being used. This behavior is what the documentation says should happen. If I upgrade the kernel which breaks the Intel HD GPU, then Plex will use the P2000 card by itself and it works fine but then I have no hardware accelerated decode streams available. nvidia-smi and vainfo both work in their respective environments.
Is there an issue on Linux where Plex will only use one GPU?
The support article says it would use the Nvidia GPU when the Intel QSV became unavailable, the way that reads to me is that the Nvidia card wouldnt be used until the Intel card was full which is where the confusion came from.
That makes sense. I disabled the intel gpu/asic so now the system only recognizes the p2000 nvidia card.
The only other issue is Plex still seems to try and use the Intel card first before trying the Nvidia card. Is there anything I can do to prevent that behavior as it does slow down the start of a stream by a few seconds ( not that its a huge burden just would be nice ) for anything that transcodes.