NVIDIA GPU troubleshooting

TrueNAS General
1

Good evening everyone. I’m having issues getting my NVIDIA GPU to be utilized by apps. According to Nvidia-SMI:

±----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.172.08 Driver Version: 570.172.08 CUDA Version: 12.8 |
|-----------------------------------------±-----------------------±---------------------+
| 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 RTX 3090 Ti Off | 00000000:44:00.0 Off | 0 |
| 0% 31C P0 88W / 450W | 0MiB / 23028MiB | 2% Default |
| | | N/A |
±----------------------------------------±-----------------------±---------------------+

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

I’ve tried removing and re-adding the nvidia drivers, and both times it appears though the nvidia drivers are working correctly. I’m fairly new with truenas usage, and I’m not really sure where to go from here. I don’t want to have to resort to a re-install if at all possible, but as long as I have a way to restore everything as it is now (except the nvidia GPU issues), I’m not completely opposed to the idea.

Thanks!

2

When you edit your apps did a checkbox apear “use this gpu” with the name of your gpu next to it? If yes did you click it? If no: click it and redeploy your app

image
image952×374 6.64 KB

1 Like
3

On each app that needs a GPU I have the box selected.

Nvidia SMI shows the following:

admin@truenas[~]$ nvidia-smi
Thu Nov 20 06:35:59 2025
±----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.172.08 Driver Version: 570.172.08 CUDA Version: 12.8 |
|-----------------------------------------±-----------------------±---------------------+
| 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 RTX 3090 Ti Off | 00000000:44:00.0 Off | 0 |
| 0% 23C P8 20W / 450W | 0MiB / 23028MiB | 0% Default |
| | | N/A |
±----------------------------------------±-----------------------±---------------------+

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

And this is what lspci shows:

admin@truenas[~]$ lspci -k | grep NVIDIA -A2
44:00.0 VGA compatible controller: NVIDIA Corporation GA102 [GeForce RTX 3090 Ti] (rev a1)
Subsystem: eVga. Corp. GA102 [GeForce RTX 3090 Ti]
Kernel driver in use: nvidia

4

When no process is using the gpu nvidia smi will show nothing

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+
root@truenas[~]# nvidia-smi
Thu Nov 20 10:08:37 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.172.08             Driver Version: 570.172.08     CUDA Version: 12.8     |
|-----------------------------------------+------------------------+----------------------+
| 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 T400 4GB                Off |   00000000:27:00.0 Off |                  N/A |
| 36%   47C    P0            N/A  /   31W |     223MiB /   4096MiB |     99%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A         3085401      C   /usr/lib/jellyfin-ffmpeg/ffmpeg         219MiB |
+-----------------------------------------------------------------------------------------+

This is how it should look like when a process, like jellyfin in my example above, uses the gpu.

5

One of the applications I have using it is NetData, and it doesn’t see any of the sensors. Immich also crashes trying to run CUDA ML tasks.

6

Reinstalled Truenas to resolve this