Hardware recommendations still up to date?

[TheDreadPirate]

Can you share the ffmpeg logs via pastebin? Also, if you use the full reply editor (click "preview post" or "new reply") you can attach images.

[fhireman]

@TheDreadPirate let me know if you need something else or if I can get those screenshots to you in another way.

Edit: ah just saw your new reply, I've added the screenshots to this reply.
how do I get the ffmpeg logs seperately? I saw some ffmpeg lines in the jellyfin docker container logs so I assumed those were the necessary ones.
do you have a command I can run with ffmpeg to try things?

[TheDreadPirate]

It looks like this is a docker container. Can you share your docker compose for the jellyfin container?

[fhireman]

I found where the ffmpeg logs were, this was the log of me trying to play something with hardware acceleration enabled: https://pastebin.com/NeYkuxhB

My compose file looks like this(using Caddy as a reverse proxy to Jellyfin):

Code:

services:
  jellyfin:
    image: jellyfin/jellyfin:latest
    container_name: jellyfin
    volumes:
      - /home/frank/jellyfin/config:/config
      - /home/frank/jellyfin/cache:/cache
      - /mnt/media:/media
    expose:
      - "8096"
    networks:
      - jellyfin-net
    restart: unless-stopped

  caddy:
    image: caddy:latest
    container_name: caddy
    ports:
      - "80:80"
    volumes:
      - ./Caddyfile:/etc/caddy/Caddyfile
    depends_on:
      - jellyfin
    networks:
      - jellyfin-net
    restart: unless-stopped

networks:
  jellyfin-net:
    driver: bridge

I'm probably missing something specific for the Jellyfin container to properly use hardware acceleration on the host machine.

[TheDreadPirate]

You did not pass the GPU into the container nor the render group ID.

https://jellyfin.org/docs/general/admini...ial-docker

Run this to get the render group ID

Code:

getent group render | cut -d: -f3

Then add this to the docker compose. Replace "122" with the value from the previous command.

Code:

group_add:
      - "122" # Change this to match your "render" host group id and remove this comment

Then pass in the GPU.

Code:

devices:
      - /dev/dri/renderD128:/dev/dri/renderD128

[fhireman]

Nice I think I got it working with:

Code:

devices:
      - /dev/dri:/dev/dri
    environment:
      - PUID=1000
      - PGID=993
    group_add:
      - "993"

ffmpeg CPU usage dropped to around 100% now and load average 0.6, way better than the 1130% and 11.2 load average before 

thanks for the help!

[fhireman]

@TheDreadPirate @crobibero

I don't quite understand, I have a lot more issues with my new setup than my old (not Jellyfin recommended) one.
old setup:
- Running v10.9.11 on Tuxedo laptop with Ubuntu 24.04.1 in Docker containers
- Ryzen 8845HS (Radeon 780M iGPU)
- 32GB RAM
- 4TB m.2 SSD that has both the OS and the media on it
- Not connected with ethernet cable, hosting on WiFi (about 200/300 Mbit on average)

Nothing done in terms of hardware acceleration, it was all CPU powered & out of the box Jellyfin container, no configuration done.
No noise issues or CPU that has high load while playing media.
Resuming an episode would not desync subtitles(or only for the first subtitle line after resuming and get back in sync at the next line).
Casting to Chromecast through Jellyfin app was flawless with no stuttering.

new setup:
- Running v10.10.2 on Ubuntu Server 24.04.1 in Docker containers
- Intel i5 12500 with UHD Graphics 770 iGPU
- 16GB RAM
- 500GB SSD for OS
- 6TB WD Red Pro for media
- connected with 1Gbit ethernet cable

High CPU usage/noise and subtitle desync issues out of the box with no configuration when resuming the same media file as on old setup.
Extra configuration to get Intel Quicksync going, lowered CPU usage/noise but still has subtitle desync issues when playing through the browser.
Casting episode from Jellyfin app to chromecast makes the episode freeze/hang for a few seconds and then play a little bit, freeze again, and play a little bit.
Resuming an episode on Jellyfin app itself has desynced subtitles.


There has to be something that I'm not configuring correctly that has a big influence on the Jellyfin setup that should be running smoother than the old AMD one. Else I'm really going to regret buying a dedicated Jellyfin Intel build


edit:
Switching subtitles in this episode from Undefined - SUBRIP- External to English - PGSSUB makes subtitles sync up again with the video, that's at least one big relief.
I'll test the casting from the jellyfin app some more this afternoon and hopefully it was just a fluke before.

[TheDreadPirate]

SRT subtitles are rendered by the browser directly, not by Jellyfin. In testing this issue in the past, Firefox was, by far, the worst offender when it came to subtitle desync.

What I noticed was that the videos that I had encoded did NOT have this issue. Other videos I had, that I did not re-encode, had desync issues sometimes. The way I was able to resolve the issue was to remux those videos and ensure that the container timestamps were correct.

Code:

/usr/lib/jellyfin-ffmpeg/ffmpeg -fflags +genpts+igndts -i "/path/to/video.mkv" -map 0 -codec copy -max_muxing_queue_size 2048 -max_interleave_delta 0 -avoid_negative_ts disabled newFile.mkv

Try that out on one of the files you are having issues with.