Arch Linux

The Troll Farmer

Member

Registered: 2025-04-01

Posts: 2

[SOLVED] black screen and no cuda after installing nvidia-dkms driver

Hi everyone. Hope you all are well.

So, I wanted to use CUDA, and since my system was using the open-source Nouveau driver, my decision was to install NVIDIA.

Following the Arch wiki NVIDIA guide, I got my card number (AD107M), Ada Lovelace, and installed the drivers for it.

First, I used the nvidia-open-dkms package because of my usage of Linux-hardened, and then the nvidia-dkms package. In both cases, my script for checking CUDA did not work, and a sighted person told me my screen was black. Not even nvidia-xconfig did anything useful.

It edited the Xorg config because I use Xorg, but I still had no screen or CUDA.

dmesg | grep display returned:
"nvidia-modeset: gpu 2.0: unable to read EDID for device dp-2"

And lspci -k -d ::03xx returned driver in use nvidia but modules were nouveau, nvidia, and nvidia_drm.

I did all the stuff about blacklisting Nouveau and all the works, but still nothing.

I reverted to a known good state by undoing all changes and reinstalling Mesa. Now, said sighted person says I have my screen back, but all outputs are still the same, and I still can't use CUDA.

The script I have is a Python one, importing PyTorch and checking if torch.cuda.is_available() is True. I also use LightDM.

Sorry if the description was not good enough; I have to work with what I have here, and I'm not good at this graphics stuff.

Have a damn good day!

Last edited by The Troll Farmer (2025-10-10 10:23:41)

---

“Today I escaped from anxiety. Or no, I discarded it, because it was within me, in my own perceptions—not outside.”

Marcus Aurelius, Meditations, 9-13

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 24,778

Re: [SOLVED] black screen and no cuda after installing nvidia-dkms driver

Eliminate linux-hardened interference and try to use a kernel with less intentional restrictions.

That said, since this is a mobile GPU chances are you should not try to use it as a primary renderer and you will want to get rid of any attempts at creating a custom xorg config nor ever run nvidia-xconfig. Just from having installed the driver packages and rebooting what happens? You'll want to look into more targeted usage of a mobile nvidia chip by looking at: https://wiki.archlinux.org/title/PRIME# … er_offload -- and invoking that specifically and only for applications were you want to use the added power of the nvidia gpu.

Should that not help, post the outputs of

sudo journalctl -b | curl -F 'file=@-' 0x0.st

Also I'm not sure whether this is a language barrier but your post is hardly legible and littered with spelling errors, you will be better understood if you take some time to clean it up -- or if you're not too familiar with english, run the text in your native language through deepl or so.