SOLVED: Hardware Acceleration

[Joshua]

Hello Jellyfin Forums

I'm trying to set up hardware acceleration. It works meaning when I try to play a video that has an unsupported format in a client it will transcode it. However the problem is it is using very high resources. My gpu will shoot up to 100%. Is this normal? I thought my gpu could handle multiple streams. I followed the documentation listed in https://jellyfin.org/docs/general/admini...tion/intel for windows. I am still experiencing this issue.


My logs:
https://pastebin.com/kSwFgT2R


specs:
Operating System: Windows 10 no docker
CPU: i512600k
GPU: Intel arc 1380

[TheDreadPirate]

Yeah. 100% usage for a single stream is normal. FFMPEG will use as much as it can get and transcoded as fast as possible. Keep in mind that it is transcoding at ~550FPS. When you start a second transcode both will transcode at the same time, but at 275FPS for each. When you start a third transcode, 184FPS. And so on.

For 1080p24 content, that speed comes out to around 20 simultaneous transcodes. 16-17 for 1080p30 content.

[Joshua]

Ah thank you for the explanation. I believe that makes sense. Wouldn't 550FPS for one stream be unnecessary though? Should or can I limit it or is that a totally normal thing to happen? Just to clarify I am hardware transcoding and not doing it by software right?

[TheDreadPirate]

You are hardware transcoding. And ffmpeg going that fast is necessary because you do want to keep ahead of where you are watching and build up a buffer. There is transcode throttling in Dashboard > Playback (near the bottom), but not in the way you are thinking. With throttling enable once ffmpeg gets far enough ahead, jellyfin will pause ffmpeg. Once the playback starts to catch up jellyfin will unpause ffmpeg and it will transcode some more at full speed.

[Joshua]

Thank you for the explanation!