[SOLVED] Nvidia driver can't find my card

AvgArchSchizoid · 2026-04-18 21:05:38

I have a brand new Arch installation on my laptop with a dual boot with windows, I downloaded the nvidia drivers from AUR to start work with cuda and cudnn.
Everything seems fine: the system boots up normally, I checked lspci and it showed that my graphics card controller is using the driver, as it should.

However, when I run nvidia-smi, it says that "No devices were found".
How does that work

My card is a Geforce RTX 5060, I'm using nvidia-beta 595.58.03-1

Forgive my dumb brain, I'm totally new to Arch

Last edited by AvgArchSchizoid (2026-04-18 22:56:39)

seth · 2026-04-18 21:09:20

it showed that my graphics card controller is using the driver

lspci -kd ::03xx

I downloaded the nvidia drivers from AUR

pacman -Qs 'nvidia|kernel'

Scimmia · 2026-04-18 21:11:42

The closed source driver does not support Blackwell cards, you have to switch to the open driver. How does it say the driver is in use, though?

AvgArchSchizoid · 2026-04-18 21:28:03

seth wrote:

lspci -kd ::03xx

02:00.0 VGA compatible controller: NVIDIA Corporation GB206M [GeForce RTX 5060 Max-Q / Mobile] (rev a1)
Subsystem: Lenovo Device 8015
Kernel driver in use: nvidia
Kernel modules: nouveau, nvidia_drm, nvidia

seth wrote:

pacman -Qs 'nvidia|kernel'

local/egl-gbm 1.1.3-1
The GBM EGL external platform library
local/egl-wayland 4:1.1.21-1
EGLStream-based Wayland external platform
local/egl-wayland2 1.0.1-1
EGLStream-based Wayland external platform (2)
local/egl-x11 1.0.5-1
NVIDIA XLib and XCB EGL Platform Library
local/libdrm 2.4.131-1
Userspace interface to kernel DRM services
local/libnefilter_conntrack 1.0.9-2
Library providing an API to the in-kernel connection tracking state table
local/libfnetlink 1.0.2-2
Low-level library for netfilter related kernel/userpace communication
local/libsysprof-capture 50.0-1
Kernel based performance profiler - capture library
local/linux 6.19.11.arch1-1
The linux kernel and modules
local/linux-api-headers 6.19-1
Kernel headers sanitize for use in userspace
local/linux-firmware-nvidia 20260309-1
Firmware files for Linux - Firmware for NVIDIA GPUs and SoCs
local/nvidia-beta 595.58.03-1
NVIDIA kernel module (beta version)
local/nvidia-utils-beta 595.58.03-1
NVIDIA drivers utilities (beta version)

Last edited by AvgArchSchizoid (2026-04-18 21:28:53)

AvgArchSchizoid · 2026-04-18 21:30:46

Scimmia wrote:

The closed source driver does not support Blackwell cards, you have to switch to the open driver. How does it say the driver is in use, though?

lspci -kd ::03xx

02:00.0 VGA compatible controller: NVIDIA Corporation GB206M [GeForce RTX 5060 Max-Q / Mobile] (rev a1)
Subsystem: Lenovo Device 8015
Kernel driver in use: nvidia <-------------- Here
Kernel modules: nouveau, nvidia_drm, nvidia

seth · 2026-04-18 21:32:36

Do you have a version of nvidia-open in the initramfs?

sudo journalctl -b | grep -i nvidia

Edit: https://www.nvidia.com/en-eu/drivers/details/265885/ actually suggests to cover the GPU

sudo journalctl -b | curl -s -H "Accept: application/json, */*" --upload-file - 'https://paste.c-net.org/'

Last edited by seth (2026-04-18 21:36:42)

AvgArchSchizoid · 2026-04-18 21:51:01

seth wrote:

Do you have a version of nvidia-open in the initramfs?

I don't think so.

sudo journalctl -b | grep -i nvidia

Apr 18 08:47:40 archlinux kernel: nvidia: loading out-of-tree module taints kernel
Apr 18 08:47:40 archlinux kernel: nvidia: module license 'NVIDIA' taints kernel.
Apr 18 08:47:40 archlinux kernel: nvidia: module verification failed: signature and/or required key missing - tainting kernel
Apr 18 08:47:40 archlinux kernel: nvidia: module license taints kernel
Apr 18 08:47:40 archlinux kernel: nvidia-nvlink: Nvlink core is being initialized, major device number 239
Apr 18 08:47:40 archlinux kernel: nvidia 0000:02:00.0: enabling devicce (0000 -> 0003)
Apr 18 08:47:40 archlinux kernel: nvidia 0000:02:00.0: vgaarb: VGA decodes changed: olddecodes=io+mem,decodes=none:owns=none
Apr 18 08:47:40 archlinux kernel: NVRM: loading NVIDIA unix X86_64 Kernel Module 595.58.03 Tue Mar 17 19:54:55 UTC 2026
Apr 18 08:47:40 archlinux systemd-modules-load[275]: Inserted module 'nvidia'
Apr 18 08:47:40 archlinux kernel: nvidia_uvm: module uses symbols nvUvmInterfaceDisableAccessCntr from proprietary module nvidia, inheriting taint.
Apr 18 08:47:40 archlinux kernel: nvidia_modeset: Loading NVIDIA Kernel Mode Setting Driver for Unix platforms 595.58.03 Tue Mar 17 19:37:50 UTC 2026
Apr 18 08:47:40 archlinux systemd-modules-load[275]: Inserted module 'nvidia_modeset'
Apr 18 08:47:40 archlinux systemd-modules-load[275]: Inserted module 'nvidia_uvm'
Apr 18 08:47:40 archlinux kernel: [drm] [nvidia-drm] [GPU ID 0x00000200] Loading Driver
Apr 18 08:47:40 archlinux kernel: NVRM: The NVIDIA GPU 0000:02:00.0 (PCI ID: 10de:2d59)
NVRM: installed in this system requires use of the NVIDIA open kernel modules.
Apr 18 08:47:40 archlinux kernel: [drm:nv_drm_dev_load [nvidia_drm]] *ERROR* [nvidia_drm] [GPU ID 0x00000200] fAILED TO ALLOCATE NvKmsKapiDevice
Apr 18 08:47:40 archlinux systemd-modules-load[275]: Inserted module 'nvidia_drm'
Apr 18 08:47:40 archlinux kernel: NVRM: The NVIDIA GPU 0000:02:00.0 (PCI ID: 10de:2d59)
NVRM: installed in this system requires use of the NVIDIA open kernel modules.
Apr 18 08:47:40 archlinux kernel: NVRM: The NVIDIA GPU 0000:02:00.0 (PCI ID: 10de:2d59)
NVRM: installed in this system requires use of the NVIDIA open kernel modules.
Apr 18 08:47:40 archlinux kernel: nvidia 0000:02:00.0: Enabling HDA Controller

Scimmia · 2026-04-18 21:52:39

seth wrote:

Edit: https://www.nvidia.com/en-eu/drivers/details/265885/ actually suggests to cover the GPU

Yes, the download includes both the closed and open drivers. You choose at install time, if there's a choice available. For Blackwell, there's not.

Last edited by Scimmia (2026-04-18 21:53:22)

AvgArchSchizoid · 2026-04-18 22:06:13

seth wrote:

sudo journalctl -b | curl -s -H "Accept: application/json, */*" --upload-file - 'https://paste.c-net.org/'

https://paste.c-net.org/DennisonEnsue

AvgArchSchizoid · 2026-04-18 22:08:55

Scimmia wrote:

seth wrote:
Edit: https://www.nvidia.com/en-eu/drivers/details/265885/ actually suggests to cover the GPU
Yes, the download includes both the closed and open drivers. You choose at install time, if there's a choice available. For Blackwell, there's not.

I tried downloading the open package before, but couldn't get it working either, it was ignoring it and using nouveau
https://aur.archlinux.org/packages/nvidia-open-beta

Scimmia · 2026-04-18 22:17:59

Then that's what we need to troubleshoot.

AvgArchSchizoid · 2026-04-18 22:54:12

I tried reinstalling it again and it just worked, still don't know why, but who am I to complain.

I guess I'm closing the thread, thank Scimmia and seth for the help and patience.

seth · 2026-04-19 06:37:43

kinda baffling that the nvidia module (correctly) rejects the GPU but still somehow binds to the device…

Please always remember to mark resolved threads by editing your initial posts subject - so others will know that there's no task left, but maybe a solution to find.
Thanks.

georgef42 · 2026-04-19 07:21:42

This actually makes sense once you look closely at what your logs are saying.

Your system is loading the proprietary NVIDIA module, but the key line is:

“requires use of the NVIDIA open kernel modules”

That basically confirms the issue — your GPU (RTX 50-series / Blackwell) is not properly supported by the closed driver, even if it appears loaded via lspci. That’s why nvidia-smi returns “No devices were found”.

What’s happening under the hood
lspci → shows kernel module attached (not full functionality)
nvidia-smi → requires full driver support → fails
Kernel logs → explicitly demand open kernel modules

So this is not a detection issue — it’s a driver compatibility mismatch.

Correct approach (for future readers)

If someone hits this again, the proper fix path is:

Remove proprietary driver:

sudo pacman -Rns nvidia-beta nvidia-utils-beta

Install open driver stack:

sudo pacman -S nvidia-open

Rebuild initramfs:

sudo mkinitcpio -P
Reboot

Also make sure nouveau is not interfering.

Important debugging tip (most people skip this)

Before assuming it's fixed, always stress test the GPU properly — because sometimes drivers “load” but fail under actual workload.

According to Arch guidance, stress testing is essential to validate stability under load, not just detection.

You can quickly test your GPU here (no install needed):
? https://devicexa.com/tools/gpu-stress-test/

Useful for:

Confirming GPU is actually being utilized
Detecting hidden driver issues
Checking stability after switching drivers

seth · 2026-04-19 07:27:06

It's not "my" log and doesn't make any sense at all - the module should™ (jaja) not bind to or at very least unbind from unsupported devices.
(There're examples of modules misbinding, but they're typically unaware that they're incapable of properly handling the device which isn't the case here)

Arch Linux

#1 2026-04-18 21:05:38

[SOLVED] Nvidia driver can't find my card

#2 2026-04-18 21:09:20

Re: [SOLVED] Nvidia driver can't find my card

#3 2026-04-18 21:11:42

Re: [SOLVED] Nvidia driver can't find my card

#4 2026-04-18 21:28:03

Re: [SOLVED] Nvidia driver can't find my card

#5 2026-04-18 21:30:46

Re: [SOLVED] Nvidia driver can't find my card

#6 2026-04-18 21:32:36

Re: [SOLVED] Nvidia driver can't find my card

#7 2026-04-18 21:51:01

Re: [SOLVED] Nvidia driver can't find my card

#8 2026-04-18 21:52:39

Re: [SOLVED] Nvidia driver can't find my card

#9 2026-04-18 22:06:13

Re: [SOLVED] Nvidia driver can't find my card

#10 2026-04-18 22:08:55

Re: [SOLVED] Nvidia driver can't find my card

#11 2026-04-18 22:17:59

Re: [SOLVED] Nvidia driver can't find my card

#12 2026-04-18 22:54:12

Re: [SOLVED] Nvidia driver can't find my card

#13 2026-04-19 06:37:43

Re: [SOLVED] Nvidia driver can't find my card

#14 2026-04-19 07:21:42

Re: [SOLVED] Nvidia driver can't find my card

#15 2026-04-19 07:27:06

Re: [SOLVED] Nvidia driver can't find my card

Board footer