Truenas scale, Plex, RTX 3050, drivers, keylase driver patch

[raudraido]

I solved my problem with re-sitting GPU from first PCI-E slot to second. Now it works as it should and everything's stable.

TrueNAS SCALE is doing the passthrough into it's own VM. I had a lot of crashing too, until I manually changed the board to run the slot at PCIe 3.0. even though the board only supports PCIe 3.0, it seems the card was still trying to negotiate 4.0. so I forced it. I've had this issue with other 30-series cards.

 

[truecharts]

On the topic of the keylase patch:
We'll try to look into doing something with this, just like our added amd support.
As it's kinda crucial for home-users.

 

[gsrcrxsi]

On the topic of the keylase patch:
We'll try to look into doing something with this, just like our added amd support.
As it's kinda crucial for home-users.

that sounds cool. thanks.

 

[boogaliwoogali]

Just to add to this, as others have mentioned patching the host TrueNAS Scale OS's Nvidia driver did indeed propagate the patch to my plex app.

To do this, I created a post-init script like so

And pointed it to a bash script containing this code

Code:

#!/bin/bash
if [ ! -d "/tmp/nvidia-patch" ]
then
    cd /tmp && git clone https://github.com/keylase/nvidia-patch.git
    cd nvidia-patch && bash ./patch.sh
else
    cd /tmp/nvidia-patch && bash ./patch.sh -s
fi

This way the patch is applied before the plex app starts deploying. Putting this here in case it helps.

 

[Sveken]

Guessing by the Jirra ticket an updated driver hasn't been implemented yet?

 

[Ppriorfl]

So Nvidia 30x0 cards still not supported? I’m trying really hard not to mess with os drivers and install non-supported patches but it’s getting painful when the 40x0 cards will be appearing soon and I’m still unable to have my card working. (3080)

 

[Sveken]

Should be all good now with the latest version, It includes 470

 

[Ppriorfl]

I installed update. Still get:

root@truenas[~]# nvidia-smi
No devices were found

Am I missing something?

 

[Sveken]

I installed update. Still get:

root@truenas[~]# nvidia-smi
No devices were found

Am I missing something?

Does transcoding not work?
The latest scale update (22.02.3) lists 470 as the current driver which Nvidia says supports 30 series.

EDIT, seems the desktop 3050 needs even newer drivers, yet the rest of the 30 series should work

 

[Fleshmauler]

Any word on when the needed drivers for the 3050 would be added?

 

[jhmw01]

Debating whether to go with the patch or what for an update, pretty crucial or home users, any news? I bought the 3050 since it was cheap, but the 3060 which is support in the support dispaly driver is like 50 quid, I would have jsut got this if id known..

 

[Fleshmauler]

Debating whether to go with the patch or what for an update, pretty crucial or home users, any news? I bought the 3050 since it was cheap, but the 3060 which is support in the support dispaly driver is like 50 quid, I would have jsut got this if id known..

I'm fairly certain that the patch simply kills nvidia's nonsense limit on non-quadro encode/decode instances.

It won't solve our woes on old drivers.

 

[jhmw01]

I'm fairly certain that the patch simply kills nvidia's nonsense limit on non-quadro encode/decode instances.

It won't solve our woes on old drivers.

Damn, I should have just bought the 3060. I'm debating sell the 3050 I have and swapping saves the headache.

 

[Fleshmauler]

[Edited] If I'm not mistaken the update on Dec 13th to SCALE 22.12.0 should also update to the nividia 515 driver which'll resolve this issue for 3050 owners.

I can't find where I read about the driver changes, but it was likely somewhere buried in https://www.truenas.com/docs/scale/scale22.12/
[/edit]

 

[jhmw01]

[Edited] If I'm not mistaken the update on Dec 13th to SCALE 22.12.0 should also update to the nividia 515 driver which'll resolve this issue for 3050 owners.

I can't find where I read about the driver changes, but it was likely somewhere buried in https://www.truenas.com/docs/scale/scale22.12/
[/edit]

I was just playing around this evening and on Plex there was hw transcoding happening remotely, a managed user. I checked by cpu which is amd so no intel quicksync and it was passing it through. The nvidia-smi shows the process too, so it must be working.

 

[jhmw01]

I was just playing around this evening and on Plex there was hw transcoding happening remotely, a managed user. I checked by cpu which is amd so no intel quicksync and it was passing it through. The nvidia-smi shows the process too, so it must be working.

nvidia-smi pmon -s u showing transcoding so yea, its working for Plex

 

[Fleshmauler]

Oh nice - are you on the RC release? If so glad to hear it is already working on it!

 

[jhmw01]

Ye

Oh nice - are you on the RC release? If so glad to hear it is already working on it!

Yea, 22.12-RC.1

 

[Fleshmauler]

Show : Issues with 3050 still not working after 22.12.01

waited for 22.12.0 - still can't see my rtx3050

*edited in confirming that I'm on driver v515.65.01

Code:

root@truenas[~]# nvidia-smi -h
NVIDIA System Management Interface -- v515.65.01

Code:

root@truenas[~]# lspci | grep VGA
29:00.0 VGA compatible controller: ASPEED Technology, Inc. ASPEED Graphics Family (rev 41)
2d:00.0 VGA compatible controller: NVIDIA Corporation Device 2507 (rev a1)

Code:

root@truenas[~]# nvidia-smi 
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Thoughts? Recommendations?

So far only thing I've tried doing is making sure the GPU is NOT isolated & rebooting.

*edit 2*
So I went ahead & saved my config file, did a clean install, uploaded my config; everything working good now.

If anyone else is having issues with a supported GPU STILL not working I'd recommend doing the same as it took maybe 10 minutes to do.

Code:

root@truenas[~]# nvidia-smi
Tue Dec 13 16:14:49 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  | 00000000:2D:00.0 Off |                  N/A |
|  0%   39C    P8     6W / 130W |      0MiB /  8192MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

 

[dasunsrule32]

Just to add to this, as others have mentioned patching the host TrueNAS Scale OS's Nvidia driver did indeed propagate the patch to my plex app.

To do this, I created a post-init script like so
View attachment 55594

And pointed it to a bash script containing this code

Code:

#!/bin/bash
if [ ! -d "/tmp/nvidia-patch" ]
then
    cd /tmp && git clone https://github.com/keylase/nvidia-patch.git
    cd nvidia-patch && bash ./patch.sh
else
    cd /tmp/nvidia-patch && bash ./patch.sh -s
fi

This way the patch is applied before the plex app starts deploying. Putting this here in case it helps.

This works perfectly on Cobia (23.10.1). Just installed my old 1080ti that I wasn't using in my old gaming rig, installed the patch and had 10 transcoded streams running and barely any gpu usage. I could probably transcode over 30 streams simultaneously with a mix of 1080p and 4k, barring any other weird driver issues. 1080ti is a beast with 2x NVENC count. Thank you!

I also added a git pull to make sure the repo is up to date as well.

Code:

#!/bin/bash
if [ ! -d "/tmp/nvidia-patch" ]
then
    cd /tmp && git clone https://github.com/keylase/nvidia-patch.git
    cd nvidia-patch && bash ./patch.sh
else
    cd /tmp/nvidia-patch && git pull && bash ./patch.sh -s
fi