Nvidia Hardware Acceleration No Longer Functioning on Plex 1.28.x

Plex Media Server
1

Server Version#: 1.28.3.6101 and 1.28.2.6151
Player Version#: 4.90.1 (Misc devices)

Nvidia GTX 1080
515.65.01
Cuda 11.7

Have consistently had issues for the last several releases of Plex Media Server on Ubuntu with Hardware Acceleration not functioning. Watching nvidia-smi and never see the job queue up. Results in either Direct playback (which is fine inside the house) or Server communication errors. Have tested driver versions 470 and 495 as well with similar results. Anyone else seeing this issue / have any recommendations?

2

what is the CPU?
When you look at /dev/dri, do you have renderD128 and renderD129?
If you have renderD129, what is the HardwareDevicePath in your Preferences.xml

Using 510.85.02 here without incident.

[chuck@glockner ~.2001]$ nvidia-smi
Sat Sep  3 14:29:15 2022       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.85.02    Driver Version: 510.85.02    CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Quadro P2200        On   | 00000000:07:00.0 Off |                  N/A |
| 57%   55C    P0    27W /  75W |   1358MiB /  5120MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

DEBUG logs which capture the failures will be needed to confirm what’s happening

3

Just D128, nothing defined in the pref file as no quicksync present. Can give that a whirl though.

If that doesn’t do it, I’ll post a debug log. Appreciate it.

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 515.65.01    Driver Version: 515.65.01    CUDA Version: 11.7     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  On   | 00000000:01:00.0  On |                  N/A |
|  8%   49C    P8     8W / 180W |      2MiB /  8192MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

by-path  card0  renderD128
4

Facepalm.

That did it. Thanks!

5

@schmidtys101, What was the fix that you used? I am also experiencing a similar issue where the latest upgrade has killed my HW transcoding.

We’ve almost the same setup, minus the driver version. I am on 510.54.

What path did you set for the HardwareDevicePath?

@ChuckPa , I only have renderD128 within /dev/dri. What logs would you need from the logs file, or would the zip of all of it be helpful for you?

6

If you make a zip of the whole Logs directory, that’s fine.
I have things setup here such that a ZIP file is easy.

Having the ZIP let’s me look back through at any other attempts which fails.

Separate quasi-related question:

Did you change the user plex runs as? If so, we need to tweak it.
If you don’t tweak it, the next package installation will do it for you :wink:

I’d consider upgrading to 510.85 unless there’s a specific reason not to.

7

Did you change the user plex runs as?

I don’t believe so. I’ve installed plex via apt just to make upgrades down the line easier - not sure if that would change anything.

But the user running plex would be cause for plex not having permissions to the GPU… either way, here’s the ZIPed logs after it failed to use my GPU for transcoding.

Thanks again for your help @ChuckPa , I’ve been banging my head against the wall for about a week now before I had the realization I could have just gone on the forums to ask for help.

8

@runrunaway

Which driver version do you have installed ?

If “bleeding edge” current I recommend backing down. I’ve seen other reports about the current

In your logs,

  1. It sees the card
  2. It cannot confirm HEVC capability (which a GT1080 does have)
Sep 09, 2022 15:20:21.544 [0x7fa00c215b00] DEBUG - [Req#4596/Transcode] [FFMPEG] - Loaded sym: cuMipmappedArrayDestroy
Sep 09, 2022 15:20:21.544 [0x7fa00c215b00] DEBUG - [Req#4596/Transcode] [FFMPEG] - Loaded sym: cuImportExternalSemaphore
Sep 09, 2022 15:20:21.544 [0x7fa00c215b00] DEBUG - [Req#4596/Transcode] [FFMPEG] - Loaded sym: cuDestroyExternalSemaphore
Sep 09, 2022 15:20:21.544 [0x7fa00c215b00] DEBUG - [Req#4596/Transcode] [FFMPEG] - Loaded sym: cuSignalExternalSemaphoresAsync
Sep 09, 2022 15:20:21.544 [0x7fa00c215b00] DEBUG - [Req#4596/Transcode] [FFMPEG] - Loaded sym: cuWaitExternalSemaphoresAsync
Sep 09, 2022 15:20:21.549 [0x7fa01427eb00] WARN - JobManager: Could not find job for handle 85092
Sep 09, 2022 15:20:21.549 [0x7fa00c215b00] ERROR - [Req#4596/Transcode] [FFMPEG] - cu->cuInit(0) failed
Sep 09, 2022 15:20:21.549 [0x7fa00c215b00] ERROR - [Req#4596/Transcode] [FFMPEG] -  -> CUDA_ERROR_UNKNOWN: unknown error
Sep 09, 2022 15:20:21.549 [0x7fa00c215b00] ERROR - [Req#4596/Transcode] [FFMPEG] - 
Sep 09, 2022 15:20:21.549 [0x7fa00c215b00] DEBUG - [Req#4596/Transcode] Codecs: hardware transcoding: opening hw device failed - probably not supported by this system, error: Unknown error occurred
Sep 09, 2022 15:20:21.549 [0x7fa00c215b00] DEBUG - [Req#4596/Transcode] Could not create hardware context for hevc
9

@ChuckPa I actually have an older version of your driver.

 nvidia-smi 
Sat Sep 10 00:49:41 2022       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.54       Driver Version: 510.54       CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:01:00.0 Off |                  N/A |
| 57%   31C    P8    14W / 180W |      1MiB /  8192MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

I’ll update to a more recent but not latest one and see if that resolves my issue. Whats funny is that my setup was working perfectly fine until I updated plex to the latest version haha

10

I don’t “jump” to a version when it comes out.

I advise reading the release announcements before upgrading and NEVER upgrading to “PlexPass” unless you’re willing to take the risk.

If the release doesn’t give you something you need — is there really any harm in waiting until the next one??

11

Sadly, updating to 515.65 did not change anything.

Here is the output to /dev/dri, which shows I only have renderD128.

root@PlexLXC ~# ls -la /dev/dri
total 0
drwxr-xr-x 3 root root        100 Sep 10 06:06 .
drwxr-xr-x 8 root root        660 Sep 10 06:11 ..
drwxr-xr-x 2 root root         80 Sep 10 06:06 by-path
crw-rw---- 1 root video  226,   0 Sep 10 06:06 card0
crw-rw---- 1 root render 226, 128 Sep 10 06:06 renderD128

Output of nvidia-smi

root@PlexLXC ~# nvidia-smi 
Sat Sep 10 06:13:50 2022       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 515.65.01    Driver Version: 515.65.01    CUDA Version: 11.7     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:01:00.0 Off |                  N/A |
| 54%   33C    P0    40W / 180W |      0MiB /  8192MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

Do you have any other clues/tips as to what I can do to fix my transcoding issue?

12

Go back down. Use the version I have. There are too many reports of issues with 515.65

[chuck@lizum ~.2012]$ gog nvidia-smi
Sat Sep 10 02:52:13 2022       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.85.02    Driver Version: 510.85.02    CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Quadro P2200        On   | 00000000:07:00.0 Off |                  N/A |
| 47%   36C    P8     4W /  75W |      1MiB /  5120MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+
[chuck@lizum ~.2013]

Your persistence isn’t configured either. Your GPU is running full power when idle.

13

@ChuckPa , thank you for all your help!

I actually tried to go back to 510.X, but that still didn’t work for me. Same issues.

Thank you for also pointing out that the persistence daemon wasn’t configured. It was actually already running, but for whatever reason, drivers 510.X and above just wasn’t working for me anymore.

I downgraded down to 495.46 and all is well again. Plex is now using my GPU for hardware transcoding.

For future users, the exact driver I used was https://us.download.nvidia.com/XFree86/Linux-x86_64/495.46/NVIDIA-Linux-x86_64-495.46.run

I am running Plex within a privileged LXC inside proxmox 7.2-7. My kernel version is 5.15.53-1-pve.

Thanks again for your help!

14

My fix was the addition of the following to my Preferences.xml

HardwareDevicePath=“/dev/dri/renderD128”

I am still running 515.65.01 for NVIDIA drivers

15

I wonder if a change in the 515.65.01 drivers now responds diferently to the nvprobe() call.

I will relay that to the transcoder team.