[RESOLVED] NVIDIA drivers failing to load eGPU on Ubuntu 22.04.1 BIOS 3.06 beta

Welcome to the community, @Alfonso_Taboada!

I believe (it’s been a bit) that nvidia-smi --query-gpu will tell you the driver in use and more about the cuda situation.

1 Like

Nvidia just released driver version 545.29.02, which appears to be the first release where the open drivers are finally beta instead of alpha.

From Chapter 44. Open Linux Kernel Modules

“Use of the open kernel modules on GeForce and Workstation GPUs should be considered Beta quality in this release and no longer requires setting of the “NVreg_OpenRmEnableUnsupportedGpus” nvidia.ko kernel module parameter. The open kernel modules are suitable for broad usage, and NVIDIA requests feedback on any issues encountered that are specific to them.”

I’m looking forward to these being packaged by Ubuntu and trying again. I never did have any luck running a display with my egpu, the various older driver/Ubuntu versions and my 12th gen Framework 13. (They would always crash fairly quickly.)

I was on Pop OS 22.04 for a long time with a working GPU after updating my BIOS to 3.06 but eventually found myself not using it anymore as there was a regression that caused my 3060 to stop working with a bunch of other issues.

I decided I very much hate regressions and have switched over to running Debian Bookworm which I am liking much more but I am still having trouble getting my drivers properly installed.

Fortunately, the display is working right out of the gate with the use of the nouveau drivers that came preinstalled and my 3060 is recognized in neofetch so that’s already better than the situation I was in. However I’m left wondering if I could see more performance benefits from installing the real drivers. I tried installing the only package, labeled nvidia-driver, that was available in the repo but was quickly back to where I started with no display and not being recognized by any utilities. Purging all the nvidia-driver gave me back a working display as well.

I do know the last time that I was successful in getting the proprietary drivers working, I had to install one of the open metapackages. However, I am left having to try to find them myself as it’s not in Debians repo. According to https://pkgs.org/download/nvidia-driver-525-open, there are 3 different versions of the metapackage, 525.147.05, 525.125.06, and 525.105.17

My question to you good people is do you know which version you are running? For further context, I am running with a 12th gen 1260P, Razer Core X Chroma, BIOS 3.06, kernel 6.1.0-13, the NVreg kernel parameter in place, GNOME with Wayland, and aforementioned Bookwork and RTX 3060

Another question I have is should I even bother trying to get the proprietary drivers running? I have a separate Windows partition for more compute heavy tasks but maybe the nouveau drivers should be good enough? At this point I can’t help but think if it ain’t broke, don’t fix it. Last thing, would switching to X11 help? I’ve read in a previous thread that forcing GDM to use X11 helped but I’ve also heard that Wayland support is getting better as well?

To determine the NVDIIA driver you’re running, you would use nvidia-smi .

Excellent question. Unlike AMD open source drivers (AMDGPU module), the nouveau are to be kind, rough. They “work” for casual tasks (video watching, browsing, writing), but you will not want to game with them or look to do anything graphically intense. It’s slowly getting better, but it’s no where even remotely as good as NVIDIA proprietary.

Has anyone tried the new open drivers, as mentioned by @Jon_Anderson?

@T8R I have installed:
ii nvidia-driver-525-open 525.147.05-0ubuntu0.22.
Works most of the time but have to reboot sometimes.

I tried just about everything I could think of to get Debian to work but still struggled to get any proprietary drivers installed correctly so I gave up and installed Ubuntu 22.04.3.

I’m happy to report that I haven’t experienced any problems except an update broke my previously working fingerprint sensor I had which I had enrolled previously but everything eGPU related is working correctly. nvidia-smi is returning with the running processes and version number. The version I have installed is 535.129.03.

It is working well and I haven’t found myself using my Windows partition as much. I haven’t stress tested it but I was able to play Portal 2 and render some basic Blender scenes quick.

Maybe one day I’ll go back to Debian when the open source drivers improve but happy to report that Ubuntu has been working out well for me