FYI Web Client still has some issues with it, can you installed the native Plex Client?
Other than that, I’m fresh out of ideas. The issues from earlier this year, to do with transcoding, seems to now be fully resolved. I can only suggest trying to install a fresh new PMS and just testing to see if it works? If there is an issue with how you’re implementing your Nvidia drivers, this could carry-over to the new PMS, but it’s still a good test to try out.
Well, I might try that but it’s a lot of work… there was a “guide” mentioning how to install drivers there Hardware HDR tonemapping still broken on 1.31.0.6654 - #200 by premikkoci
But by looking how to passthrough GPU to a docker instance, I also seen that at least on Ubuntu, there is dedicated headless drivers. Is there any equivalent on RHEL? Maybe it worth the try…
Another question: is there anything I can setup in my preferences file, I’ve seem the option “VaapiDriver=XXXX” is existing.
@ChuckPa if you have any answer to give me… 
Do you have any experience to tell me which Nvidia drivers would be the “best” to choose (branch, version, repository…) to install it on a dedicated AlmaLinux 8 server? I’m planning a dedicated server (but still not sure if Plex will run bare metal or on docker on this box)
But about drivers, more I dig, more I’m confused to chose which ones: “Production” version numbers seems higher than “Functionality” ones. For me it doesn’t makes sense! 
Thx in advance
I use the 525.105.17 drivers
Here’s the unfiltered list from dpkg for my machine
[chuck@glockner ~.1997]$ dpkg -l | grep nvidia
ii gpustat 0.6.0-1 all pretty nvidia device monitor
ii libnvidia-cfg1-525-server:amd64 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA binary OpenGL/GLX configuration library
ii libnvidia-common-525-server 525.105.17-0ubuntu0.22.04.1 all Shared files used by the NVIDIA libraries
ii libnvidia-compute-525-server:amd64 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA libcompute package
ii libnvidia-decode-525-server:amd64 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA Video Decoding runtime libraries
ii libnvidia-egl-wayland1:amd64 1:1.1.9-1.1 amd64 Wayland EGL External Platform library -- shared library
ii libnvidia-encode-525-server:amd64 525.105.17-0ubuntu0.22.04.1 amd64 NVENC Video Encoding runtime library
ii libnvidia-extra-525-server:amd64 525.105.17-0ubuntu0.22.04.1 amd64 Extra libraries for the NVIDIA Server Driver
ii libnvidia-fbc1-525-server:amd64 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA OpenGL-based Framebuffer Capture runtime library
ii libnvidia-gl-525-server:amd64 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA OpenGL/GLX/EGL/GLES GLVND libraries and Vulkan ICD
ii libnvidia-ml-dev:amd64 11.5.50~11.5.1-1ubuntu1 amd64 NVIDIA Management Library (NVML) development files
ii nvidia-compute-utils-525-server 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA compute utilities
ii nvidia-cuda-dev:amd64 11.5.1-1ubuntu1 amd64 NVIDIA CUDA development files
ii nvidia-cuda-gdb 11.5.114~11.5.1-1ubuntu1 amd64 NVIDIA CUDA Debugger (GDB)
ii nvidia-cuda-toolkit 11.5.1-1ubuntu1 amd64 NVIDIA CUDA development toolkit
ii nvidia-cuda-toolkit-doc 11.5.1-1ubuntu1 all NVIDIA CUDA and OpenCL documentation
ii nvidia-dkms-525-server 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA DKMS package
ii nvidia-driver-525-server 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA Server Driver metapackage
ii nvidia-kernel-common-525-server 525.105.17-0ubuntu0.22.04.1 amd64 Shared files used with the kernel module
ii nvidia-kernel-source-525-server 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA kernel source package
ii nvidia-opencl-dev:amd64 11.5.1-1ubuntu1 amd64 NVIDIA OpenCL development files
ii nvidia-prime 0.8.17.1 all Tools to enable NVIDIA's Prime
ii nvidia-profiler 11.5.114~11.5.1-1ubuntu1 amd64 NVIDIA Profiler for CUDA and OpenCL
ii nvidia-utils-525-server 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA Server Driver support binaries
ii nvidia-visual-profiler 11.5.114~11.5.1-1ubuntu1 amd64 NVIDIA Visual Profiler for CUDA and OpenCL
ii xserver-xorg-video-nvidia-525-server 525.105.17-0ubuntu0.22.04.1 amd64 NVIDIA binary Xorg driver
[chuck@glockner ~.1998]$
As I write this, Ubuntu is updating to -
reconfiguring packages ...
(Reading database ... 180783 files and directories currently installed.)
Preparing to unpack .../00-ubuntu-drivers-common_1%3a0.9.6.2~0.22.04.4_amd64.deb ...
Unpacking ubuntu-drivers-common (1:0.9.6.2~0.22.04.4) over (1:0.9.6.2~0.22.04.3) ...
Preparing to unpack .../01-libmm-glib0_1.20.0-1~ubuntu22.04.2_amd64.deb ...
Unpacking libmm-glib0:amd64 (1.20.0-1~ubuntu22.04.2) over (1.20.0-1~ubuntu22.04.1) ...
Preparing to unpack .../02-nvidia-driver-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking nvidia-driver-525-server (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../03-libnvidia-extra-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking libnvidia-extra-525-server:amd64 (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../04-libnvidia-common-525-server_525.125.06-0ubuntu0.22.04.1_all.deb ...
Unpacking libnvidia-common-525-server (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../05-libnvidia-gl-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking libnvidia-gl-525-server:amd64 (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../06-nvidia-dkms-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Removing all DKMS Modules
Done.
Unpacking nvidia-dkms-525-server (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../07-nvidia-kernel-source-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking nvidia-kernel-source-525-server (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../08-nvidia-kernel-common-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking nvidia-kernel-common-525-server (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../09-libnvidia-decode-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking libnvidia-decode-525-server:amd64 (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../10-libnvidia-compute-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking libnvidia-compute-525-server:amd64 (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../11-nvidia-compute-utils-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking nvidia-compute-utils-525-server (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../12-libnvidia-encode-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking libnvidia-encode-525-server:amd64 (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../13-nvidia-utils-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking nvidia-utils-525-server (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../14-xserver-xorg-video-nvidia-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking xserver-xorg-video-nvidia-525-server (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../15-libnvidia-fbc1-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking libnvidia-fbc1-525-server:amd64 (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../16-libnvidia-cfg1-525-server_525.125.06-0ubuntu0.22.04.1_amd64.deb ...
Unpacking libnvidia-cfg1-525-server:amd64 (525.125.06-0ubuntu0.22.04.1) over (525.105.17-0ubuntu0.22.04.1) ...
Preparing to unpack .../17-modemmanager_1.20.0-1~ubuntu22.04.2_amd64.deb ...
Unpacking modemmanager (1.20.0-1~ubuntu22.04.2) over (1.20.0-1~ubuntu22.04.1) ...
Setting up libnvidia-fbc1-525-server:amd64 (525.125.06-0ubuntu0.22.04.1) ...
Setting up nvidia-kernel-common-525-server (525.125.06-0ubuntu0.22.04.1) ...
I will report in a moment
EDIT: Updating to 525.125.06 (Ubuntu vetted) works correctly for a P2200
I’m on the same version still as @ChuckPa, being 525.105.17. I’ve never used AlmaLinux, but am using Ubuntu 20.04. So I’d just stick with that if you can. I did mess with 530 drivers, when I had a docker+gpu issue, but I just reverted back to 525 as they’re working well; they’re also the latest stable “server” edition/version (I think 
).
I’m using ESXi with h/w passthrough (not really ideal and kinda frowned upon 
) and haven’t had any issues from ESXi/VMs. I did recently spin up a docker container in said VM, just so I could have a 2nd instance of Plex that could use the GPU, and this actually works really well. It took me a while to get it going though. And again, not idea going from ESXi, to VM, to Docker, to the final container 
I’d suggest just playing around with your build/*nix to try and get it working, and also find it’s limitations. Basically try and break it Maximum streams + maximum transcodes. Throw some subtitles at it (as that caused loads of problems). I did this with my builds so I can find any weaknesses… This gives you plenty of times to re-install and try other solutions; there is nothing worse than rolling out a new PMS and trying to use it in prod, only for it to have issues!
The thing that is not clear to me, is do I need to install CUDA “on the top” of nvidia drivers for Plex and if yes, which version? For now, I’ve installed 525 series with CUDA 12.0 (12.1 and 12.2 didn’t want to install for dependencies reasons).
BTW, is there something else to install for decoding/transcoding properly (codecs)?
But this “mess” is really annoying because you don’t really know what will work and what will not unless trying!
I’m using ESXi with h/w passthrough (not really ideal and kinda frowned upon
Yes I know the feeling because that is what I had (and I will try now docker on a physical machine because I have other plans for that machine like home assistant so “passthrough to Docker only” might be my solution for a while if I can validate the GPU tests. This also gives a nice start base to introduce all the ARRs containers later on…
I’m since 3 months in a “Docker” binge and migrated almost all my web apps - Like Netbox, Leantime, Grafana, etc… in this new direction, Plex should follow!
I use Ubuntu SERVER for all this because the Nvidia drivers are vetted and stable. ( I’m not stable so I need all the help I can get 
) Running a desktop environment in a VM is a complete waste.
Plex already exists in Docker and has for a LONG time.
– No, we’re not going to binge because our bellies are full 
LOL
Plex w/ HW transcoding under ESXi
– Start Preparing VM definition
– Allocate memory for the VM
– RESERVE the memory (lock VMs memory into RAM – required step)
– ADD PCI device (either QSV - UHD device or Nvidia)
– Add ISO to boot and create VM
Load up guest VM
Confirm packages libnvidia-encode and libnvidia-decode were also installed
– manually install these if needed (apt-get install …)
Confirm nvidia-smi sees the card.
Install Plex.
I know you don’t want to hear this … 
Well, I think I DID IT !!! 
Almalinux Server Headless, Docker and GPU passthrough.
I could validate the HW transcoding on a test instance so I did the swap from the VM to the container.
It works VERY good (the host is a I5, 32 GB RAM, 1 TB SSD and 2x10GB/s in LACP for the network. I could run 4x4K streams without an issue (after patching of course). Web client works finally again.
But install drivers is a f***** pain on Linux, I hesitated to go back to an headless Windows core !!!
The only “loss” for now is that I had a reverse proxy on the VM itself to use the web client on port 443, which is now back to 32400. I could not find a way to have both unless using another reverse proxy somewhere else. Is there a way to have internal port 32400 available on 2 external ports with docker?
I think I’ll close the thread but I have to say it: it was, despite all the technical challenges, a real pleasure exchanging with you guys! 
This is why I use Ubuntu Server cough cough
driver install was as simple as “sudo apt install nvidia-drivers-525-server”
Supported stuff “just works”
me neither! ROFL
Glad you persevered and got it working 
Couldn’t event get Cockpit to work on it properly without creating a “fake” bond… (it’s my favorite remote control tool for Linux)
Well, I tihink Linux is all about compromises… today I struggled get the proper backup solution for my physical server.
Could not get Veeam agent working, neither Synology agent… so I went with a basic rsync to backup all my containers files! My goodness it was painful!!! 
Back at that time, I “translated” the Plex Windows registry settings to the Preference.xml file… I didn’t want to do that backwards, but if I want to go back to VM, maybe “this is the way” to get NVIDIA properly passthrough to Plex! 
Well, we all have different mental health troubles I guess…
…This is the way …
When the Nvidia drivers are installed on the host or when using QSV,
Putting PMS , with HW transcoding, in the LXC container is:
From the command line:
lxc config device add CONTAINER_NAME gpu gpu gid=$(getent group $(ls -la /dev/dri | grep renderD128 | awk '{print $4}') | awk -F: '{print $3}')
The process does the following:
To test:
root@glockner:~# echo lxc config device add CONTAINER_NAME gpu gpu gid=$(getent group $(ls -la /dev/dri | grep renderD128 | awk ‘{print $4}’) | awk -F: ‘{print $3}’)
lxc config device add CONTAINER_NAME gpu gpu gid=109
root@glockner:~#
Now restart the container (or LXDware VM)
Ensure PMS is a member of the GID/group so the passthrough works.
ps: had to break out my light saber to remember how 
_
If I’m using docker (which I loathe)
sudo docker run \
-d \
--name plex \
--network=host \
-e PLEX_CLAIM="claim-xxxxxx" \
-e TZ="EST" \
-e LANG="en_US.UTF-8" \
-e PLEX_UID=1000 \
-e PLEX_GID=GID_OF_DEV_DRI \
-e PUID=1000 \
-e PGID=GID_OF_DEV_DRI \
-h dockerplex \
-v /sata/dockerplex:/config \
-v /sata/dockerplex/tmp:/tmp \
-v /sata/dockerplex/transcode:/transcode \
-v /vie:/data \
-v /glock:/glock \
--device=/dev/dri:/dev/dri \
plexinc/pms-docker:plexpass
#docker start plex
#docker update --restart=unless-stopped plex
Should we be scared?
(That’s a lot of transcoding stuff…)
HW transcoding is not working here after upgrading PMS to 1.32.5.7328
Using nvidia driver 535.54.03 on Debian. PMS running in docker. HW transcoding works fine on 1.32.4.7195-7c8f9d3b6 and earlier versions.
Same issue and log messages both when I select the GPU or pick Auto
Snippet from logs when starting playback.
Jul 21, 2023 22:40:20.152 [139781555485496] DEBUG - [Req#111d/Transcode] Codecs: testing h264_nvenc (encoder)
Jul 21, 2023 22:40:20.152 [139781555485496] DEBUG - [Req#111d/Transcode] Codecs: hardware transcoding: testing API nvenc for device 'pci:0000:06:1b.0' (GA104GL [RTX A4000])
Jul 21, 2023 22:40:20.152 [139781555485496] DEBUG - [Req#111d/Transcode] [FFMPEG] - Loaded lib: libcuda.so.1
Jul 21, 2023 22:40:20.152 [139781555485496] DEBUG - [Req#111d/Transcode] [FFMPEG] - Loaded sym: cuInit
Jul 21, 2023 22:40:20.152 [139781555485496] DEBUG - [Req#111d/Transcode] [FFMPEG] - Loaded sym: cuDeviceGetCount
...
Jul 21, 2023 22:40:20.153 [139781555485496] DEBUG - [Req#111d/Transcode] [FFMPEG] - Loaded sym: cuEGLStreamProducerPresentFrame
Jul 21, 2023 22:40:20.153 [139781555485496] DEBUG - [Req#111d/Transcode] [FFMPEG] - Loaded sym: cuEGLStreamProducerReturnFrame
Jul 21, 2023 22:40:20.153 [139781555485496] ERROR - [Req#111d/Transcode] [FFMPEG] - cu->cuDeviceGetByPCIBusId(&hwctx->internal->cuda_device, pci_bus_id) failed
Jul 21, 2023 22:40:20.153 [139781555485496] ERROR - [Req#111d/Transcode] [FFMPEG] - -> CUDA_ERROR_INVALID_DEVICE: invalid device ordinal
Jul 21, 2023 22:40:20.153 [139781555485496] ERROR - [Req#111d/Transcode] [FFMPEG] -
Jul 21, 2023 22:40:20.153 [139781555485496] DEBUG - [Req#111d/Transcode] Codecs: hardware transcoding: opening hw device failed - probably not supported by this system, error: Generic error in an external library
lspci shows:
06:1b.0 VGA compatible controller: NVIDIA Corporation GA104GL [RTX A4000] (rev a1)
I need more than that snippet.
Please include the entire the log file which shows from the MDE → FFMPEG switchover
Here you go
transcode.log (140.2 KB)
This is a mess.
Jul 22, 2023 02:16:38.937 [139793322867512] DEBUG - [Req#59/Transcode] [FFMPEG] - Loaded sym: cuEGLStreamProducerDisconnect
Jul 22, 2023 02:16:38.937 [139793322867512] DEBUG - [Req#59/Transcode] [FFMPEG] - Loaded sym: cuEGLStreamConsumerDisconnect
Jul 22, 2023 02:16:38.937 [139793322867512] DEBUG - [Req#59/Transcode] [FFMPEG] - Loaded sym: cuEGLStreamProducerPresentFrame
Jul 22, 2023 02:16:38.937 [139793322867512] DEBUG - [Req#59/Transcode] [FFMPEG] - Loaded sym: cuEGLStreamProducerReturnFrame
Jul 22, 2023 02:16:38.937 [139793322867512] ERROR - [Req#59/Transcode] [FFMPEG] - cu->cuDeviceGetByPCIBusId(&hwctx->internal->cuda_device, pci_bus_id) failed
Jul 22, 2023 02:16:38.937 [139793322867512] ERROR - [Req#59/Transcode] [FFMPEG] - -> CUDA_ERROR_INVALID_DEVICE: invalid device ordinal
Jul 22, 2023 02:16:38.937 [139793322867512] ERROR - [Req#59/Transcode] [FFMPEG] -
Jul 22, 2023 02:16:38.937 [139793322867512] DEBUG - [Req#59/Transcode] Codecs: hardware transcoding: opening hw device failed - probably not supported by this system, error: Generic error in an external library
Jul 22, 2023 02:16:38.937 [139793322867512] DEBUG - [Req#59/Transcode] Could not create hardware context for hevc
Jul 22, 2023 02:16:38.937 [139793322867512] DEBUG - [Req#59/Transcode] Once Upon a Time in the City of New York - audio.channels limitation applies: 6 > 2
Jul 22, 2023 02:16:38.937 [139793322867512] DEBUG - [Req#59/Transcode] MDE: Cannot direct stream audio stream due to profile or setting limitations
Jul 22, 2023 02:16:38.937 [139793322867512] DEBUG - [Req#59/Transcode] MDE: Beauty and the Beast - S1 E1 - Once Upon a Time in the City of New York: selected me
Is this the 535 driver ? It looks like it based on the error.
Is this in a VM?
I’ve spent the last 2+ hours trying to install 22.04.2 LTS on AlderLake with Nvidia 3040 GPU and it hangs every time the INSTANT I boot up after installing the 525 or 535 driver.
I have 535 on my Xeon machine (P2200) and it’s solid.
Yes the host machine is running Proxmox 8 with Debian 12 VM + GPU passthrough (RTX A4000) and 535.54.03 driver. Running smooth & stable until the latest PMS update. Was on Proxmox 7.4 + 525 before that, running for ~6 months without a single issue.
I am running this on Threadripper PRO. I have had no luck running PMS stable on Xeon machines.
