When it doesn’t see the hardware /dev/dri inside the container, the log will look like this. (It starts looking for external GPUs)
which CPU are you using?
When it doesn’t see the hardware /dev/dri inside the container, the log will look like this. (It starts looking for external GPUs)
which CPU are you using?
CPU is Intel J3455 (running a synology DS918+)
/dev/dri is there, right permissions are set.
I started downgrading until I found that latest version where my hw transcoding works is :
plexmediaserver_1.32.1.6999-91e1e2e2c_amd64.deb
Thanks.
Thx @mitzone i upgrade my docker Host to bookworm, and i the hardware accélération stop working, i used to blamed debian, but it’s just a bug in Plex
I revert m’y docker-compose to plexinc/pms-docker:1.32.1.6999-91e1e2e2c
Know i have hardware accélération ![]()
In the beta, you’ll find the ApolloLake fix (1.32.4)
Thanks for your reply.
I just installed the beta version (1.32.4.7195-7c8f9d3b6 is this the one?)
HW transcoding still no go.
root@PlexServer:/# ls -al /dev/dri/
total 0
drwxr-xr-x 2 root root 80 Jun 13 22:12 .
drwxr-xr-x 6 root root 380 Jun 13 22:12 …
crw-rw-rw- 1 root video 226, 0 Jun 13 22:12 card0
crw-rw-rw- 1 root video 226, 128 Jun 13 22:12 renderD128
root@PlexServer:/# id plex
uid=1037(plex) gid=100(users) groups=100(users),65536(video),138862(PhotoStation)
errors in log
Jun 13, 2023 22:12:33.393 [139896273566520] WARN - [Req#cb] SLOW QUERY: It took 2900.000000 ms to retrieve 6 items.
Jun 13, 2023 22:12:35.298 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Failed to initialise VAAPI connection: -1 (unknown libva error).
Jun 13, 2023 22:12:35.299 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Cannot load libcuda.so.1
Jun 13, 2023 22:12:35.299 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Could not dynamically load CUDA
Jun 13, 2023 22:12:35.300 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Failed to initialise VAAPI connection: -1 (unknown libva error).
Jun 13, 2023 22:12:35.301 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Cannot load libcuda.so.1
Jun 13, 2023 22:12:35.301 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Could not dynamically load CUDA
Jun 13, 2023 22:12:35.301 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Cannot load libcuda.so.1
Jun 13, 2023 22:12:35.301 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Could not dynamically load CUDA
Jun 13, 2023 22:12:35.302 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Failed to initialise VAAPI connection: -1 (unknown libva error).
Jun 13, 2023 22:12:35.303 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Cannot load libcuda.so.1
Jun 13, 2023 22:12:35.303 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Could not dynamically load CUDA
Jun 13, 2023 22:12:35.303 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Cannot load libcuda.so.1
Jun 13, 2023 22:12:35.304 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Could not dynamically load CUDA
Jun 13, 2023 22:12:35.304 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Failed to initialise VAAPI connection: -1 (unknown libva error).
Jun 13, 2023 22:12:35.305 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Cannot load libcuda.so.1
Jun 13, 2023 22:12:35.305 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Could not dynamically load CUDA
Jun 13, 2023 22:12:35.306 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Cannot load libcuda.so.1
Jun 13, 2023 22:12:35.306 [139896324381496] ERROR - [Req#f5/Transcode] [FFMPEG] - Could not dynamically load CUDA
back to 1.32.1.6999-91e1e2e2c
I have been posting around the forum.
I must apologize.
The ApolloLake fixes did not make it into 1.32.4 prior to being submitted to QA.
They WILL be in 1.32.5
This unfortunately means you need to roll back again to what you had working until next release.
Got it, thanks.
Will wait for 1.32.5
I have the same problem, only plexinc/pms-docker:1.32.1.6999-91e1e2e2c works running on QNAP TS-435B with latest QTS
It’s the same problem because the TS-435B is also apollolake CPU
Intel® Celeron® J3455 4-core 1.5 GHz processor (burst up to 2.3 GHz)
Any word on when 1.32.5 will be released? I thought it was supposed to be Jun 19th?
Releases are planned but if there’s QA rework then it will be delayed to get it right.
I know we’re all waiting for 1.32.5
1.32.5.7210 beta now available.
FIXES:
4K HW transcoding with HDR tone mapping still doesn’t work. At least HW transcoding works on non-HDR videos again with my J3455 system. A step in the right direction.
I still have the same issue with the latest stable and plex pass PMS releases. I tried 1.32.5.7328 and 1.32.5.7349 with no luck – HW decoding and encoding not working. As soon as I downgrade to 1.32.1.6999 HW transcoding works again without issue.
System Specs:
OS: Bare metal Ubuntu 22.04 (22.04.6 LTS)
Kernel: 6.3.7-060307-generic (previously 5.4.0 before PMS broke HW transcoding)
CPU: i5-8500T (Coffee Lake)
PMS: Bare metal install (no container or VM)
Any thoughts on what the issue may be here?
Please verify DEBUG logging is enabled. VERBOSE disabled.
Recreate the failure.
Download and attach the server logs ZIP file.
CoffeeLake CPUs should not be failing on 22.04
Afterthought: Do you have VaapiDriver="i965" in your Preferences.xml file?
definitely not crazy about pushing the kernel version above distro standard
Nailed it on the last point. I had VaapiDriver=“i965” in Preferences.xml along with a manual backup dated March 2021 to get this machine HW transcoding originally. As soon as I removed it and restarted, HW transcoding is working on the latest PMS.
Very happy to hear that the new drivers have been incorporated and working for “newer” CPUs. Thanks for helping and getting this one solved so quickly.
And agreed on the Kernel. I was very happy with the original 5.4.0, granted that one aged a bit over the years, but I was at my wits end originally trying to get HW transcoding working again a few months back. I’ll probably leave it for now until the distro catches up
I’m running 22.04.2 on a 5.15 kernel (it upgraded me during the update)
I’m very happy sitting where I am
Holy… thanks Chuck! I was on a kernel version 6.3.6 with my J3455. I downgraded to version 5.15 and all was good again… finally! This issue was driving me nuts!
When you play on the
edge, you’ll invariable need ![]()
Recommend staying with mainstream kernels for production systems.
![]()
That’s great advice. When I upgraded to Ubuntu 22.04.3 LTS, the kernel installed was newer and ever since then it broke HW transcoding with tonemapping. A few Plex beta versions ago, HW transcoding started to work again, but tonemapping was still broken. Reverting back to kernel version 5.15 fixed all my transcoding issues.