nicknsy / jellyscrub

Smooth mouse-over video scrubbing previews for Jellyfin.
MIT License
660 stars 25 forks source link

HW Acceleration disabled despite plugin saying its enabled #146

Open zzggbb opened 4 months ago

zzggbb commented 4 months ago

In the jellyscrub plugin settings, I have full hardware acceleration enabled. In my playback settings I have Nvidia hardware acceleration enabled, for all codecs.

The jellyscrub plugin used to run ffmpeg with CUDA, but it randomly stopped using CUDA. I have tried restarting Jellyfin and my computer. Why would the plugin not use HW accel if the plugin setting has it enabled?

GGLinnk commented 4 months ago

I think I'm facing the same issue. I was monitoring my server because it was slow, until I discovered that ffmpeg was running for generating bif files. I've run a command to check my GPU stats it wasn't used. But it is used when transcoding media. I haven't been able to start HW E/D yet...

deadcxap commented 3 months ago

Similar problem. Processor i5-12100, hardware encoding configured. When watching movies, everything is fine, I see through intel_gpu_top that the video core is loaded and used for encoding. When using jellyscrub, for the first 10-30 minutes after activating the plugin, it works using hardware encoding, and after that it starts to load only the processor, without using the video core.

nicknsy commented 3 months ago

Are you guys doing HW acceleration through some virtualization layer like docker? Also, if you force a video to transcode, for instance by lowering the bitrate during video playback, do you see that your GPU is being correctly used by Jellyfin even when it is not being used by Jellyscrub?

GGLinnk commented 3 months ago

Are you guys doing HW acceleration through some virtualization layer like docker? Also, if you force a video to transcode, for instance by lowering the bitrate during video playback, do you see that your GPU is being correctly used by Jellyfin even when it is not being used by Jellyscrub?

Yes, I'm using a docker container. I've tested the nvidia-smi command (inside the container) and it worked great and movies are transcoded. There is only the bug which display that there is transcodes but is not listing them explicitly. But that's another thing. Only jellyscrub is using the CPU.

I must say that these test were done a while now, I've not checked if the situation where different today.