[Support] ich777 - Jellyfin AMD/Intel/Nvidia [DEPRECATED]

[anylettuce]

not to push it, 10.8.0 is officially out

 

[ich777]

8 hours ago, anylettuce said:

not to push it, 10.8.0 is officially out

 

Yes I know, already saw that the build yesterday failed on my side.

Will try to fix this today.

 

Fixed and updated.

[anylettuce]

2 hours ago, ich777 said:

Yes I know, already saw that the build yesterday failed on my side.

Will try to fix this today.

 

Fixed and updated.

Thanks, I hope this fixes a bug that I hate.

[ich777]

29 minutes ago, anylettuce said:

Thanks, I hope this fixes a bug that I hate.

They changed the name from ffmpeg that‘s why it failed.

To fix this was actually on my todo list but currently I‘m really busy and I had planed to fix this today evening if nobody requests it…

[JudMeherg]

7 hours ago, ich777 said:

They changed the name from ffmpeg that‘s why it failed.

To fix this was actually on my todo list but currently I‘m really busy and I had planed to fix this today evening if nobody requests it…

 

Came here to check on this.

 

So thankful for devs like you that have the skills to help us normal people have awesome things!!

 

Can't wait to give this new update a shot. Seems like it has some really good changes.

[anylettuce]

18 hours ago, ich777 said:

They changed the name from ffmpeg that‘s why it failed.

To fix this was actually on my todo list but currently I‘m really busy and I had planed to fix this today evening if nobody requests it…

 

It was nothing with you that I needed fixed. If you had multiple versions of the same movie, you couldn't easily group them. Was a bug that put the TMDB id into the IMDB id. Its fixed now

[domrockt]

omg iam so sorry i diddnt look up this thread for a while!!! 

 

for AMD IGPU transcoding you all need to put 

1)      /dev/dri/renderD128 inside the docker template INSTEAD just /dev/dri

2)    /dev/dri/renderD128 in the Jellifin Gui under playback INSTEAD just /dev/dri

 

 

[ich777]

14 minutes ago, domrockt said:

omg iam so sorry i diddnt look up this thread for a while!!! 

 

for AMD IGPU transcoding you all need to put 

1)      /dev/dri/renderD128 inside the docker template INSTEAD just /dev/dri

2)    /dev/dri/renderD128 in the Jellifin Gui under playback INSTEAD just /dev/dri

That's not entirely true you can of course let everything as it is with /dev/dri in the template:

 

and in the container it will automatically put in the Jellyfin app /dev/dri/renderD128 (this is also the common way to do it - if you have changed it once than it is maybe set to /dev/dri instead to /dev/dri/renderD128) :

 

If you do it this way it will also ensure that if you have more GPUs that are capable of VAAPI on the host that you can for example switch in the container from one to another card (eg: /dev/dri/renderD128 or /dev/dri/renderD129 and so on...).

 

BTW: You should be also able to use Quick Sync if you have Intel 8th gen+ in my container.

[cthulhu161]

Yesterday I updated Unraid OS and Jellyfin and suddenly I have a strange playback issue. Everytime I watch a video it works for at least half the length without any problems. Then it suddenly stops and I’m back at the home screen. The progress is saved so (if the problem doesn’t appear in the last minutes) I can jump back to the point where I was kicked out.

 

The problem appears independent of the length of videos, the player (browser, jellyfin desktop) and the device I’m using. I tried to increase encoding speed, but it doesn’t solve anything.

 

Has anybody any idea what could be the cause of this problem?

[ich777]

1 hour ago, cthulhu161 said:

The problem appears independent of the length of videos, the player (browser, jellyfin desktop) and the device I’m using. I tried to increase encoding speed, but it doesn’t solve anything.

Is it possible that your drive where the transcodes or better speaking the cache is is running full?

[cthulhu161]

5 minutes ago, ich777 said:

Is it possible that your drive where the transcodes or better speaking the cache is is running full?

I don't think so. I have more than enough free storage place on all drives. The appdata share has 210GB free space. The only possible storage problem could be if I limited the cache - but I don't remember doing so or where such a setting would be

 

Looking into the logs I found a few possible errors: The first message that could be relevant and appears is 'Current HLS implementation doesn't support non-keyframe breaks but one is requested, ignoring that request'

Then multiple messages appear: 'Slow HTTP Response from X to Y in Z with Status Code 200' (X is a URL for Jellyfin, Y an IP, and Z is always a different number - maybe a timer?)

Then the first real error [ERR] appears: 'Jellyfin.Server.Middleware.ExceptionMiddleware: Error processing request: "Could not find a part of the path '/null'". URL "GET" "/Environment/DirectoryContents".'

Then another warning: 'Emby.Server.Implementations.HttpServer.WebSocketConnection: WS "192.168.0.178" error receiving data: "The remote party closed the WebSocket connection without completing the close handshake."'

Then: 'Jellyfin.Server.Middleware.ResponseTimeMiddleware: Slow HTTP Response from "http://192.168.0.206:8096/Sessions/Playing/Stopped" to "192.168.0.160" in 0:00:01.7688115 with Status Code 204' and after that the previous warning about the failed handshake again. Resulting in: "Emby.Server.Implementations.HttpServer.WebSocketManager: WS "192.168.0.178" closed"

 

Then the log states 'Emby.Server.Implementations.HttpServer.WebSocketManager: WS "192.168.0.178" request' and many of the previous warnings and errors reappear. This process results in 'Jellyfin.Api.Helpers.TranscodingJobHelper: Stopping ffmpeg process with q command for "/config/transcodes/676f3ed03c0245c7a27e0ef55a7281fd.m3u8"' and 'ellyfin.Api.Helpers.TranscodingJobHelper: FFmpeg exited with code 0'

 

Sometimes it also states: 'Emby.Server.Implementations.Session.SessionWebSocketListener: Sending ForceKeepAlive message to 1 inactive WebSockets.' and then 'Emby.Server.Implementations.Session.SessionWebSocketListener: Lost 1 WebSockets.' directly followed by the ffmpeg code 0 message

 

[cthulhu161]

Searching for similar errors I found that people who have multiple jellyfin/emby/plex/etc-dockers experience similar problems. I only got one docker of this kind. But many months ago I had another problem with jellyfin that could be resolved by reinstalling it. The new jellyfin couldn't find my old appdata because it created a new folder 'Jellyfin-AMD-Intel-Nvidia' and my old appdata was in a folder simply called 'Jellyfin'. This problem could be resolved by changing the Data Path in the Unraid Settings. Since then I have two folders 'Jellyfin' and 'Jellyfin-AMD-Intel-Nvidia'. Is it possible that this causes my problems?

[cthulhu161]

Tried a lot of things and nothing seemed to work. Reinstalling, changing settings, etc. - always the same problem. Changed the repository to amd64_stable_10.7.7 and this finally solved the problem. The 10.8 Upgrade doesn't seem to work on my Unraid at the moment

[ich777]

19 hours ago, cthulhu161 said:

Searching for similar errors I found that people who have multiple jellyfin/emby/plex/etc-dockers experience similar problems. I only got one docker of this kind.

Can you point me to those threads?

I mostly only use Emby/Jellyfin and never got such an issue...

 

On what clients have you tried it so far?

 

7 hours ago, cthulhu161 said:

Changed the repository to amd64_stable_10.7.7 and this finally solved the problem. The 10.8 Upgrade doesn't seem to work on my Unraid at the moment

Tried it now and let a file play in the background and got no issues at all, about 1:15 playtime so far (with HW transcoding).

[fffighter]

Hi I have few question regarding intel transcoding setting.

My unraid system running on Intel 12900k and 3090 GPU.

For some reason, I cant enable intel qsv or va-api for transcoding. Even gpu statics plug in cant detect intel igpu stat.

Is anyway I can enable qsv or va-api transcoding?

 

Also, I can enable Nvidia transcoding, but when transcoding 4k to 4k, my cpu running around 60% and my 3090 running maybe 60% too.

Is that normal??

 

I have been working on those issues for last two days, But I can't find solution, hope I can find my answer.

 

Thanks

 

 

 

Edited June 19, 2022 by fffighter

[ich777]

15 minutes ago, fffighter said:

For some reason, I cant enable intel qsv or va-api for transcoding. Even gpu statics plug in cant detect intel igpu stat.

Is anyway I can enable qsv or va-api transcoding?

Have you installed the Intel GPU Top plugin and blacklisted your iGPU in the first place?

 

You also have to pass through the device /dev/dri to the container.

 

16 minutes ago, fffighter said:

Also, I can enable Nvidia transcoding, but when transcoding 4k to 4k, my cpu running around 60% and my 3090 running maybe 60% too.

Are you talking about transcoding 4k (h265) to 4k (h264), if yes, then I have to say that h264 isn't really designed to transcode to 4k

 

Also don't forget that your CPU does transcoding from audio, so it may be possible that this is the case in your instance, also subtitles sometimes make a huge difference in terms of transcoding.

 

The next thing to also consider is that throttling is broken in Jellyfin from what I know and it makes no difference if you enable or disable it, it will always transcode at full blast.

 

 

Maybe try to transcode your 4k footage to something lower like 1080p to start and see how the utilization is then.

Have you only tried one file so far or did you try different files? What about 1080p (h264) footage, is it there the same?

[fffighter]

1 hour ago, ich777 said:

Have you installed the Intel GPU Top plugin and blacklisted your iGPU in the first place?

-- I did installed intel gpu top plugin.

-- For the blacklisted ipgu, I am not so sure how to do it, can you explain?

You also have to pass through the device /dev/dri to the container.

-- I did this one followed instruction 

 

1 hour ago, ich777 said:

 

Are you talking about transcoding 4k (h265) to 4k (h264), if yes, then I have to say that h264 isn't really designed to transcode to 4k

-- I am not sure about that, I gonna double check. Only thing I am sure the sources file is hevc 10bit 4k

 

Also don't forget that your CPU does transcoding from audio, so it may be possible that this is the case in your instance, also subtitles sometimes make a huge difference in terms of transcoding.

-- possible.

 

1 hour ago, ich777 said:

The next thing to also consider is that throttling is broken in Jellyfin from what I know and it makes no difference if you enable or disable it, it will always transcode at full blast.

--ok, this is the reason, I want to use intel igpu because the power usage. 

 

 

Maybe try to transcode your 4k footage to something lower like 1080p to start and see how the utilization is then.

-- I will try it.

Have you only tried one file so far or did you try different files? What about 1080p (h264) footage, is it there the same?

-- I will try this too.

 

 

Thank you for your super fast reply. I am in the middle of reinstall unraid server. I will try your suggestion asap

 

Sorry I forgot. If I did not install Nvidia gpu driver. Gpu stastic plug in can detect intel igup.   But the plug in did'nt pick up any live data. The display simply blank, no data shows.

 

 

Did more digging, It looks I need to wait unraid use new Linux kernel 5.16 or later to fix intel gpu issue.

 

Edited June 19, 2022 by fffighter
add information.

[ich777]

13 hours ago, fffighter said:

-- For the blacklisted ipgu, I am not so sure how to do it, can you explain?

Do this from a Unraid terminal:

echo "blacklist i915" > /boot/config/modprobe.d/i915.conf

and reboot the Server.

[fffighter]

6 hours ago, ich777 said:

Do this from a Unraid terminal:

echo "blacklist i915" > /boot/config/modprobe.d/i915.conf

and reboot the Server.

 

Awesome, after I blacklist i915,  It looks works both on qsv or vaapi. cpu load pretty much lower if I chose low power mode compare with using nvidia.

Hdr works too, the movie color not pale anymore unlike use cpu only decode.

 

Thank you so much

 

The new issue is the subtitles. When I turn on the subtitles, it display in the middle of picture.

 

change subtitle position in the setting  won't work. anyway to fix it?

 

The other issue is, in the libraries view, the cast and crew photo wont display. After click the cast name, it will show up. It is really annoying.  Anyway to fix it?

 

Edited June 20, 2022 by fffighter
add information.

[ich777]

14 minutes ago, fffighter said:

The new issue is the subtitles. When I turn on the subtitles, it display in the middle of picture.

This is a thing from Jellyfin and can be changed in the menu, if not please create a dedicated issue for that on GitHub if you don't work it out because that's out of my control and should be reported to the developers from Jellyfin itself.

 

15 minutes ago, fffighter said:

The other issue is, in the libraries view, the cast and crew photo wont display. After click the cast name, it will show up. It is really annoying.  Anyway to fix it?

This is also a thing that you should report on their GitHub:

 

Click

[9to5crypto]

Hello! After the update to Jellyfin version 10.8.0 I was hoping I could finally use OpenCL tone mapping but I cannot make it work. I'm currently using Intel QSV with a UHD 630 iGPU and transcoding with tone mapping makes ffmpeg crash with code 139. Does anyone have the same problem?

[ich777]

34 minutes ago, 9to5crypto said:

Hello! After the update to Jellyfin version 10.8.0 I was hoping I could finally use OpenCL tone mapping but I cannot make it work. I'm currently using Intel QSV with a UHD 630 iGPU and transcoding with tone mapping makes ffmpeg crash with code 139. Does anyone have the same problem?

With that little information I can't help any further, please post your Diagnostics.

[G0ku737]

Hey there,

 

i have a strange problem that appered recently (last jellyfin update maybe):

 

I cant play "MPEG-4 Visual (XviD)" encoded videos. 

For that part not so strange..but..it worked before..the exact same files.

Thats not all. In the Jellyfin App on my Pixel 6 Pro i can watch them no problem.

In the browser i get the message: "This client isn't compatible with the media and the server isn't sending a compatible media format." and on my Shield TV it just dont starts playing showing : "to many errors...".

 

Also what is weird: when i watch one of the files on the Pixel the GPU usage is 0% while watching.

When i watch a 4k h.264/5 the gpu is utilization is as normal 12-15%.

 

When i deactivate HW acceleration i can play every file just normal but with much higher CPU load.

Even when i watch the same file as before on my Pixel 6 Pro the CPU usage is 20-30% higher that with HWA activated which does not make sense to me because the GPU utilization before was 0%.

 

I hope someone can help me. I can provide any data necessary to solve the issue (if you tell me where to get it).

 

Thanks 

 

[ich777]

17 minutes ago, G0ku737 said:

I hope someone can help me. I can provide any data necessary to solve the issue (if you tell me where to get it).

Please try to use VAAPI instead of QSV if you are using QSV

[9to5crypto]

1 hour ago, ich777 said:

With that little information I can't help any further, please post your Diagnostics.

You're right, sorry about that. Here's the diagnotics zip file and a pastebin of Jellyfin's log file. diagnostics-20220622-2020.zip