Getting HW Transcoding running

[Handling6959]

Hello everybody,

I'm currently trying to get my completely redesigned homelab setup up and running.

For this I use an HPE DL380 Gen9 with two powerful Xeon E5-2690 v4 CPUs.

Sure, this HW would be powerful enough to do the transcoding on the CPU, but I also got an Nvidia Tesla P4 GPU.

I use a Proxmox Hypervisor and the Jellyfin Server VM runs on Windows Server 2022 with 16 threads and 32GB RAM.
The GPU is correctly recognized in Windows Device Manager with the latest official drivers from Nvidia.
Driver Version: 31.0.15.3815

Now I thought, I'm ready to activate HW transcoding. In Jellyfin I have activated the "Nvidia NVENC" hardware decoding.
Unfortunately, I don't notice any difference when playing a wide variety of videos. The CPU load increases up to 80% for 4k and the transcoded files are correctly created in the corresponding temporary folder. But I assume that the Nvidia Tesla P4 GPU is not used and everything looks behaves exactly the same with HW transcoding deactivated.

1. How can I debug the situation in more detail? Is my GPU HW and driver compatible? In the taskmanager I don't see the performance graph for the GPU but on the other hand the device manager looks fine.

2. Maybe the WS 2022 OS makes problems in combination with GPU support? Shall I switch to a Win10 VM?

3. Does GPU transcoding even make sense in my setup or should I just stick with CPU transcoding? There would be enough computing power.

Thanks!

[TheDreadPirate]

You would need to share your ffmpeg log with us.

C:\ProgramData\Jellyfin\Server\log