Jellyfin & Quicksync Issues

I have some potential issues that I am looking for confirmation/resolution on.

The first is whether or not I setup my jellyfin container to correctly utilize Quicksync. Within the docker section for jellyfin, when I first configured all of this months ago, for some reason I added both of these lines to the extra parameters section. 

--device /dev/dri/renderD128:/dev/dri/renderD128
--device /dev/dri/card0:/dev/dri/card0

Looking back at the instructions for linuxserver.io's version of this jellyfin container, it seems only --device=/dev/dri:/dev/dri should be included, as described in the Hardware Acceleration section of https://github.com/linuxserver/docker-jellyfin

Restarting the container to pass in the entire /dev/dri folder, then starting something in jellyfin that requires transcoding for the client, running intel_gpu_top within a console of the unraid host pc results in something similar to the following:
 

intel-gpu-top: Intel Alderlake_s (Gen12) @ /dev/dri/card0 - 1257/1443 MHz;   0% RC6;  2.92/45.62 W;    19769 irqs/s

         ENGINES     BUSY                                                                                                          MI_SEMA MI_WAIT
       Render/3D   71.84% |█████████████████████████████████████████████████████████████████████████▍                            |      0%      0%
         Blitter    0.00% |                                                                                                      |      0%      0%
           Video   25.13% |█████████████████████████▊                                                                            |      0%      0%
    VideoEnhance    0.00% |                                                                                                      |      0%      0%

   PID              NAME           Render/3D                      Blitter                        Video                      VideoEnhance           
 19689            ffmpeg |█████████████████▊          ||                            ||█████▋                      ||                            |

They key part that had me worried that I set it up incorrectly is the fact that it lists /dev/dri/card0 instead of /dev/dri/renderD128. I would imagine renderD128 should be listed here, because thats what Jellyfin is utilizing to transcode. 

If I go to the logs of jellyfin, the ffmpeg command to transcode the content looks like this:

[2023-07-08 10:33:07.636 -04:00] [INF] [16] Jellyfin.Api.Helpers.TranscodingJobHelper: "/usr/lib/jellyfin-ffmpeg/ffmpeg" "-analyzeduration 200M -init_hw_device vaapi=va:,driver=iHD,kernel_driver=i915 -init_hw_device qsv=qs@va -filter_hw_device qs -hwaccel qsv -hwaccel_output_format qsv -c:v h264_qsv -autorotate 0 -i file:<redacted> -autoscale 0 -map_metadata -1 -map_chapters -1 -threads 0 -map 0:0 -map 0:1 -map -0:s -codec:v:0 h264_qsv -preset 7 -look_ahead 0 -b:v 292000 -maxrate 292000 -bufsize 584000 -profile:v:0 high -level 41 -g:v:0 72 -keyint_min:v:0 72 -vf \"setparams=color_primaries=bt709:color_trc=bt709:colorspace=bt709,scale_qsv=w=426:h=284:format=nv12\" -codec:a:0 libfdk_aac -ac 2 -ab 128000 -af \"volume=2\" -copyts -avoid_negative_ts disabled -max_muxing_queue_size 2048 -f hls -max_delay 5000000 -hls_time 3 -hls_segment_type mpegts -start_number 0 -hls_segment_filename \"/config/data/transcodes/af3ec090ed41953b9acce436bf7903c3%d.ts\" -hls_playlist_type vod -hls_list_size 0 -y \"/config/data/transcodes/af3ec090ed41953b9acce436bf7903c3.m3u8\""

I am assuming that the inclusion of

Quote

-hwaccel qsv

confirms that it is indeed utilizing quicksync?

Still curious why intel_gpu_top lists /dev/dri/card0 though.

For the second issue, I am trying to understand exactly what is happening with the following. I have 4k content, and if I decide to say set the quality in jellyfin to something ridiculous, like 4k @ 120Mbps the cpu struggles, resulting in big slow downs on playback, especially after seeking. I noticed a pattern that during the slowdowns, intel_gpu_top will look like this:

intel-gpu-top: Intel Alderlake_s (Gen12) @ /dev/dri/card0 - 1246/1446 MHz;   0% RC6;  1.08/19.48 W;        0 irqs/s

         ENGINES     BUSY                                                                                                          MI_SEMA MI_WAIT
       Render/3D    0.00% |                                                                                                      |      0%      0%
         Blitter    0.00% |                                                                                                      |      0%      0%
           Video   50.00% |███████████████████████████████████████████████████▏                                                  |     50%      0%
    VideoEnhance    0.00% |                                                                                                      |      0%      0%

   PID              NAME           Render/3D                      Blitter                        Video                      VideoEnhance           
 24614            ffmpeg |                            ||                            ||██████████████▉             ||                            |

The video portion is just capped at 50%, during the entirety of the playback being paused due to not transcoding quick enough. According to jellyfin, this Video section represens the QSV decoder/encoder workload, as described in Jellyfin's documentation here: https://jellyfin.org/docs/general/administration/hardware-acceleration/intel/#verify-on-linux

Any ideas why it would be capped at 50% and utilizing 100%?

Lastly, shouldnt I expect the VideoEnhance section to be utilized somewhat? My Jellyfin playback seciton looks like this:
 

Motherboard: ASUS TUF GAMING Z690PLUS WF D4

CPU: Intel i5-12600k

Thanks!

Edited July 8, 20232 yr by halexh