GPU not found

TrueNAS General
1

I’ve just installed Truenas Scale 24.04, and I see the error messages cycling on the screen:

NVRM: No NVIDIA GPU found. 
nvidia-link: Unregistered Nvlink Cire, major device number 240
nvidia-link: Nvlink Core is being initialized, major device number 240
NVRM: The NVIDIA GeForve GTX 650 Ti GPU installed in this system 
NVRM: is supported through the NVIDIA 470.xx Legacy drivers. 
NVRM: Please visit www.nvidia.com/object/unix.html for more information. 
NVRM:  The 545.23.08 NVIDIA driver will ignore this GPU. Continue probe...

I’m guessing I have to downgrade the NVIDIA driver to the supported one 470.xx Legacy. How to do that?

Thanks
Larry

2

Hi @larrybg

Unfortunately we don’t have a supported method to downgrade the NVIDIA drivers to the Legacy branch. There might be a way to do it manually but it’s likely to cause overall instability and conflicts with upgrading.

Is it possible to pick up a newer, supported card like the Quadro P400 if you’re just looking for video transcode capabilities?

3

Is that the only option? I didn’t have problems with this h/w configuration on Truenas Core.

4

If you aren’t planning to use the more advanced functionality of your NVIDIA card (video encode/decode, GPU-compute, etc) then there’s no harm in only using the base display driver - however, a 650TI is probably not as power-efficient as alternative or newer cards.

5

To be honest all I need is the Plex server and Immich. I agree that my card isn’t powerful, I’m just reusing the old Dell computer. I need to check if it can support anything better than the installed card. Its an old Dell XPS

6

So, if I leave it as it is right now, is there a way to stop that cycling message on the screen? It looks like the system is constantly trying to engage that driver with the current card

7

I’ve only ever seen that message come up once during boot-time in dmesg for unsupported/legacy cards, not cycling through the console. Perhaps it’s Plex trying to constantly poll/restart the NVIDIA services.

Barring really weird systems, PCI Express tends to just work, and is backwards compatible. NVIDIA tends to include legacy BIOS/CSM support in their GPUs, and a P400 isn’t that new.

8

I have the same “issue”
Old PC that I’m just playing with… not critical at all… getting basically the same message as the OP… and I saw the question asked but no answer…
the console keeps cycling the message over and over…
is there a way to stop it?
Or is it an issue at all and just let it go and ignore the message?

This is with an old Nvidia 210…

9

Anyone? Just keeps cycling… is my only option to remove the old GPU?

10

I’m getting what I think is the same issue using the current build of Dragonfly (24.4.02?) using a K600 gpu. I know my gpu is not supported with the 545.23.0 driver included in this version. I know the last version to support the 470.x driver is Angelfish (original/22.x) release. I also know upgrading from 22.x to 23.x drops the legacy driver (expected, but still a bummer). I don’t need to use the gpu other than for the console (if I do later I will just get a newer card and install it as a second gpu).

What I would like to do is suppress the console messages as they are distracting at a minimum, and rather intrusive when doing console activities.

I suspect there is a service or config that could be disabled via the CLI to stop this behavior (beyond me though). Likely whatever is calling for NVLINK CORE. Possibly Kubernetes/Containers.

  • I THINK these started after activating support for Applications.
  • I THINK the message appears once per network port. I have 6 network ports on my machine and I get 5 of these dumps every 5 minutes. Only one port is currently connected. I will try disabling the other ports to see if that changes anything (pessimistic).

I do plan to install some apps later, so I’d rather not break or disable support for apps and virtual machines trying to resolve this.

Actually, there are other services I’d like to disable too to optimize the boot (eg: I am not running truenas under vmware so those things could be disabled, etc, but not an issue for today).

11

Update: I counted and got six of these NVLINK messages, so that matches 1:1 with the six network ports on my machine.

So far disabling some interfaces has had interesting results (not the good kind). Seems like Truenas really does not want users to disable physical interfaces once it finds them.

12

I’m having the same issue as well except I’m not using Plex or any other video app either.

13

I think I will go back on Core. Used it for about 3 years without any problems, why I even bothered to try Scale.
The Scale is unstable and has performance issues, sometimes I have to restart the computer because I cannot login or GUI is unresponsive.