Can’t install Nvidia drivers

Can’t install nvidia drivers

Hello everybody,

I’m ex Debian user and I’m trying to install nvidia drivers on my opensuse tumbleweed for my nvidia rtx4060ti.

In the following documentation https://en.opensuse.org/SDB:NVIDIA\_drivers it clearly says that first I need to run zypper install openSUSE-repos-Tumbleweed-NVIDIA (which for my surprise, it uninstalls the microos version that randomly got installed on my computer). And after that I run zypper in nvidia-open-driver-G07-signed-kmp-meta .

The issue is that:

  1. MOKutils doesn’t show up in the terminal asking for a password. Also there is no nvidia-pubkeys folder in /usr/share
  2. The nvidia driver that is installed has only 120MB.
  3. Resolution is locked at 768p and 60hz (my monitor is 244hz).

I can’t find any official documentation about installing nvidia drivers. I want the full nvidia driver experince like I’ve had it on Debian.

How did you managed to work it out with it?

Is this documentation obsolete?

Thanks!

Welcome (again :wink: ). Please open a terminal and issue:

su -
<password>
zypper lr -d
zypper se -si nvidia

Then copy and paste the output of the two last lines, including the command and the trailing prompt, between code tags (use the </> button above the editing area).

I’ve just reinstalles another OS because I couldn’t use my computer with tumbleweed. But to answer to your question:

  1. I had download.nvidia.com/opensuse/tumbleweed repo
  2. nvidia-open-driver-G07-signed-kmp-meta + opensuse-repos-Tumbleweed-NVIDIA packages

Hope that this little info can help you figuring out what happened on my system.

Thank you

There is currently a mismatch with the drivers in the Nvidia repo (see the reply to your other post for details).
At this point, the easiest way is to wait a week or two before installing Tumbleweed again.

1 Like

I think Version mismatch between the kmp and the others…

But that is looking in a glass…

But when i run “zypper in nvidia-open-driver-G07-signed-kmp-meta” it supposed to download other things too? It only downloaded a 120mb driver. A fully nvidia driver has over 600mb

As already written, there is currently a mismatch in available drivers; once you reinstall Tumbleweed, post the actual output of the commands and we will help you through that. The SDB:Nvidia_drivers is current and correct, the repos are not at the time of writing this.

this issue with desync drivers is common or i just had really bad luck?

also, if i had opensuse leap installed, i would have the same issue?

Bad luck, holiday times, openSUSE → Nvidia sync etc…

No, Leap 16.0 is fine, I do see the nvidia driver in weak-updates after a kernel update, but that’s not a version jump like on Tumbleweed.

I’m using the open driver from the cuda repository which uses dkms here (no 32bit) it’s working fine.

Does dkms solve nvidia mismatch problem? If so why it’s not installed by default?

No.

openSUSE has it own mechanism to get the drivers build when the kernel gets updated.

But now, the open-nvidia-kmp is updated to a higher version of nvidia and the nvidia packages from nvidia not.

This is a problem how they are build.

Tumbleweed and nvidia Repo will have sometimes problems.

Why @malcolmlewis keeps saying dkms can solve nvidia kernel version mismatch … :weary:

@VIVADE The Nvidia driver from the cuda repository rebuilds/signs etc with dkms, not the one from the oss repository which is kernel specific.

You mean people who use proprietary drivers can benefit from dkms but not those who use open drivers. Is that the idea?

@VIVADE I’m using the open driver, the default is to use that if your GPU is Turing architecture or better.

The cuda repository does not have any 32bit support.

Edit: Add dkms info

journalctl -W -b -11 | grep dkms

Aug 13 08:37:11 dkms[804]: Sign command: /lib/modules/7.1.8-1-default/build/scripts/sign-file
Aug 13 08:37:11 dkms[804]: Signing key: /var/lib/dkms/mok.key
Aug 13 08:37:11 dkms[804]: Public certificate (MOK): /var/lib/dkms/mok.pub
Aug 13 08:37:11 dkms[804]: Autoinstall of module nvidia/610.43.02 for kernel 7.1.8-1-default (x86_64)
Aug 13 08:37:13 dkms[1297]: 'make' -j4 KERNEL_UNAME=7.1.8-1-default IGNORE_PREEMPT_RT_PRESENCE=1 IGNORE_XEN_PRESENCE=1 modules
Aug 13 08:40:25 dkms[1297]:  done.
Aug 13 08:40:25 dkms[1297]: Signing module /var/lib/dkms/nvidia/610.43.02/build/kernel-open/nvidia.ko
Aug 13 08:40:25 dkms[1297]: Signing module /var/lib/dkms/nvidia/610.43.02/build/kernel-open/nvidia-modeset.ko
Aug 13 08:40:25 dkms[1297]: Signing module /var/lib/dkms/nvidia/610.43.02/build/kernel-open/nvidia-drm.ko
Aug 13 08:40:26 dkms[1297]: Signing module /var/lib/dkms/nvidia/610.43.02/build/kernel-open/nvidia-uvm.ko
Aug 13 08:40:26 dkms[1297]: Signing module /var/lib/dkms/nvidia/610.43.02/build/kernel-open/nvidia-peermem.ko
Aug 13 08:40:27 dkms[1297]: Installing /lib/modules/7.1.8-1-default/updates/nvidia.ko.zst
Aug 13 08:40:27 dkms[1297]: Installing /lib/modules/7.1.8-1-default/updates/nvidia-modeset.ko.zst
Aug 13 08:40:27 dkms[1297]: Installing /lib/modules/7.1.8-1-default/updates/nvidia-drm.ko.zst
Aug 13 08:40:27 dkms[1297]: Installing /lib/modules/7.1.8-1-default/updates/nvidia-uvm.ko.zst
Aug 13 08:40:27 dkms[1297]: Installing /lib/modules/7.1.8-1-default/updates/nvidia-peermem.ko.zst
Aug 13 08:40:27 dkms[1297]: do_depmod 7.1.8-1-default
Aug 13 08:40:32 dkms[1297]:  done.
Aug 13 08:40:35 dkms[804]: Autoinstall on 7.1.8-1-default succeeded for module(s) nvidia.

I’m more confused now. Never mind but I won’t install dkms as I don’t know what is its purpose if it doesn’t solve the problem.

1 Like

@VIVADE AFAIK for the NVIDIA:repo-non-free (Tumbleweed and Leap 16.0) there is no nvidia type dkms service, only in the cuda repository.

Information for package dkms:
-----------------------------
Repository     : cuda
Name           : dkms
Version        : 3.4.1-1
Arch           : noarch
Vendor         : NVIDIA
Installed Size : 183.5 KiB
Installed      : Yes (automatically)
Status         : up-to-date
Source package : dkms-3.4.1-1.src
Upstream URL   : https://github.com/dell/dkms
Summary        : Dynamic Kernel Module Support Framework
Description    : 
    This package contains the framework for the Dynamic Kernel Module Support (DKMS)
    method for installing module RPMS as originally developed by Dell.

I’m on a Dell workstation with a RTX4000.

AFAIK, it should work on any system…

… and the CUDA repo offers the “New Feature Branch” (currently 610.57.04) which is aimed at early adopters / developers / power users and might have its own snags, so not for the casual user and not a good candidate for default install if you don’t know what you are doing.
The OSS package usually lags the Nvidia one by a week or two and since the Nvidia repo keeps the old versions you should not see the specific problem described in this thread, but the CUDA driver is not so carefree as it appears in the expert hands of Malcolm :wink:

It does lag, the current cuda driver is 610.43.02 :wink:

1 Like

I see nvidia-open-driver-G07-signed-cuda-kmp-default-610.57.04_k7.1.8 in the OSS repo
Time for a zypper dup, or am I missing something?