Nvidia GPU users attention please! [June 2025]

chuckie · June 19, 2025, 6:04pm

Hi team! I hope everyone is having a great day. Today, I updated my system with eos-update --aur and restarted my machine. Then, I try to run my game but it wasn’t running at all, so I did a little bit of research and found these:

Running nvidia-smi:

Failed to initialize NVML: Driver/library version mismatch
NVML library version: 575.64

Then checking on journalctl -b:

kernel: NVRM: API mismatch: the client 'nvidia-smi' (pid 35618)
NVRM: has the version 575.64, but this kernel module has
NVRM: the version 575.57.08.  Please make sure that this
NVRM: kernel module and all NVIDIA driver components
NVRM: have the same version.

I would like to know if anyone face the same problem and how to fix it, thank you for your attention in advance!

Cphusion · June 19, 2025, 6:27pm

Share the output of this?

joekamprad · June 20, 2025, 11:47am

packages all have to be the same version .. 575.64-1
one of nvidia modules packages and nvidia-utils package.

but this kernel module has
NVRM: the version 575.57.08

so your modules packagfe is not updated or you have the initramfs image not rebuilded ..

chuckie · June 20, 2025, 2:27pm

╰─➤  pacman -Qs nvidia
local/egl-gbm 1.1.2.1-1
    The GBM EGL external platform library
local/egl-wayland 4:1.1.19-1
    EGLStream-based Wayland external platform
local/egl-x11 1.0.2-1
    NVIDIA XLib and XCB EGL Platform Library
local/lib32-nvidia-utils 575.64-2
    NVIDIA drivers utilities (32-bit)
local/libnvidia-container 1.17.8-1
    NVIDIA container runtime library
local/libvdpau 1.5-3
    Nvidia VDPAU library
local/nvidia 575.64-1
    NVIDIA kernel modules
local/nvidia-container-toolkit 1.17.8-1
    NVIDIA container toolkit
local/nvidia-hook 1.5.2-1
    pacman hook for nvidia
local/nvidia-inst 25.6.5-1
    Script to install/uninstall nvidia driver packages in EndeavourOS
local/nvidia-lts 1:575.64-2
    NVIDIA drivers for linux-lts
local/nvidia-utils 575.64-1
    NVIDIA drivers utilities

Thank you for your comments, recently there was an update on EndeavourOS that fix that issue. I want to know about this one rebuild initramfs image, I will investigate but if you have some info where I can start learning about it, I would appreciated! Again, thank you very much for you help, have a great weekend!

Cphusion · June 20, 2025, 2:34pm

You can run sudo dracut-rebuild to rebuild your initramfs.

j0rbsh · June 20, 2025, 3:32pm

I was getting the same error about mismatch version for NVML. In my case when I tried dracut-rebuild it failed because my /efi parttion was full. I had to remove some old folders and then dracut succeeded and after a reboot I was in working order again.

joekamprad · June 20, 2025, 7:17pm

thats one case that could happen, if the regeneration fails, or does not happen caused by issue .. the initramfs can include still old modules for nvidia.

in case of Grub Bootloader it’s that and in case of default systemd-boot Bootloader it is sudo reinstall-kernels command

Cphusion · June 20, 2025, 7:39pm

Thanks I don’t know all the EndeavourOS tools good to learn of the difference between the two. I’ll try to remember those.

joekamprad · June 20, 2025, 8:29pm

in general they will run automatically in case needed a rebuild we have so called hooks that does this for you. only in some case you may need/want to make sure thy rebuild like after manual changes.

read here for more info:
https://discovery.endeavouros.com/installation/dracut/2022/12/

bob · June 21, 2025, 2:06pm

Running an eos instance in virtualbox here. After rebooting following the latest yay update there is a failure to launch the gui. Tried the fix in the first post, didn’t work. Each subsequent boot stalls just before launching the gui. Have rolled back to a working state for now.

In answer to an earlier post, here are some commands with their output (I stress that the commands are run prior to the update that borks my system):

pacman -Q | grep -P 'nvidia|nouveau'
No output

lspci -vnn | grep -A9 -P 'VGA|3D|Display'
00:02.0 VGA compatible controller [0300]: VMware SVGA II Adapter [15ad:0405] (prog-if 00 [VGA controller])
        Subsystem: VMware SVGA II Adapter [15ad:0405]
        Flags: bus master, fast devsel, latency 64, IRQ 18
        I/O ports at d010 [size=16]
        Memory at e0000000 (32-bit, prefetchable) [size=128M]
        Memory at f0000000 (32-bit, non-prefetchable) [size=2M]
        Expansion ROM at 000c0000 [virtual] [disabled] [size=128K]
        Kernel driver in use: vmwgfx
        Kernel modules: vmwgfx

00:03.0 Ethernet controller [0200]: Advanced Micro Devices, Inc. [AMD] 79C97x [PCnet32 LANCE] [1022:2000] (rev 40)

lsmod | grep -P 'nvidia|nouveau'
No output

Is this related? I have an nvidia card.

Cphusion · June 21, 2025, 3:34pm

No because your vm has virtual vmware video adapter.

joekamprad · June 21, 2025, 9:37pm

As @Cphusion mentioned, as long as you do not pass your Nvidia to the VM by intent it will not use it directly and instead using virtual GPU and driver.

joekamprad · June 21, 2025, 11:18pm

Spreadneck · June 22, 2025, 8:36pm

About to do a reinstall, does this still apply?

joekamprad · June 24, 2025, 10:09am

ISO is hot fixed, you should not have the issue as long as you have internet connection [for new installs].

NeedaVirus · June 24, 2025, 5:42pm

Had the issue, (of course…)
Workaround worked great for me

Many thanks !

jdjohnston · June 26, 2025, 10:10pm

Honest, I didn’t forget. Capturing the output from commands in the console was a bit tricky. /usr/bin/scrtipt might have been helpful, but I forgot about it. In any case, I did get the results, but when I went to post the results last week, the EOS website was down.

In any case, combining the instructions from you & Der Doktor, I found:
`pacman -Qs nvidia
local/egl-gbm 1.1.2.1-1
The GBM EGL external platform library
local/egl-wayland 4:1.1.19-1
EGLStream-based Wayland external platform
local/egl-x11 1.0.2-1
NVIDIA XLib and XCB EGL Platform Library
local/libvdpau 1.5-3
Nvidia VDPAU library
local/nvidia-470xx-dkms 470.256.02-7.98
NVIDIA drivers - module sources
local/nvidia-470xx-utils 470.256.02-7.98
NVIDIA drivers utilities

----- couldn’t remove libvdpau (dependencies) -----

pacman -Qs nvidia
local/libvdpau 1.5-3
Nvidia VDPAU library

ls /etc/X11/xorg.conf.d/
00-keyboard.conf

NB. I added “algol” & “cdrom” to the following grep patterns to make sure the search was
empty & not just something I did wrong. Empty results sometimes make me nervous.

pacman -Q | grep -P ‘nvidia|nouveau|algol’
algol68g 3.5.12-1
algol68g-doc 3.5.12-1

lspci -vnn | grep -A9 -P ‘VGA|3D|Display’
00:02.0 VGA compatible controller [0300]: VMware SVGA II Adapter [15ad:0405] (prog-if 00 [VGA controller])
Subsystem: VMware SVGA II Adapter [15ad:0405]
Flags: bus master, fast devsel, latency 64, IRQ 18
I/O ports at d010 [size=16]
Memory at e0000000 (32-bit, prefetchable) [size=64M]
Memory at f0000000 (32-bit, non-prefetchable) [size=2M]
Expansion ROM at 000c0000 [virtual] [disabled] [size=128K]
Kernel driver in use: vmwgfx
Kernel modules: vmwgfx

00:03.0 Ethernet controller [0200]: Intel Corporation 82540EM Gigabit Ethernet Controller [8086:100e] (rev 02)

lsmod | grep -P ‘nvidia|nouveau|cdrom’
cdrom 81920 1 sr_mod
`
Does this give any clue? Remember, I’m running EOS as a VirtualBox VM on an old Dell laptop with Windows 7 as the host OS.

Thanks for the ongoing advice

jdjohnston · June 26, 2025, 10:13pm

Please see my belated response to manuel’s request for comand output where I combined your instructions with his.

Thanks for the previous advice.

manuel · June 27, 2025, 7:47am

Are you trying to use Nvidia driver in a VM?

joekamprad · June 27, 2025, 8:09am

If you do not use GPU PCI passthrough with the Nvidia GPU you do not need Nvidia drivers in your VM it will use virtual devices