Installed Alma 10.1…fantastic, then, install NVIDIA drivers, using Wiki, and now, video stucked on lowest resolution… how to come back to the good NVIDIA driver ? Thanks.
Did you install the right driver for your card? You may want to run “nvidia-detect” provided by elrepo:
And try installing elrepo’s nvidia driver from there.
Hello and welcome to the forum.
Before I can help, I need a little more information about your hardware:
First, run this command to find out which video devices and module are assigned:
for v_var1 in $(lspci | egrep 3D\|VGA | grep NVIDIA | cut -d’ ’ -f1); do lspci -ks ${v_var1};done
Example:
Second: The wiki shows two ways to install NVIDIA drivers:
open
and proprietary
Which one did you install?
Important: The open driver primarily supports Turing (2018) and newer architectures.
Regards!
First, thank you for sharing your time an knowledge.
Here is the results of the command
01:00.0 VGA compatible controller: NVIDIA Corporation AD104 [GeForce RTX 4070] (rev a1)
Subsystem: Gigabyte Technology Co., Ltd Device 40ee
Kernel driver in use: nvidia
Kernel modules: nouveau, nvidia_drm, nvidia
I have been using the first method (open), and probably the second one after discovering the issue.
Here is another set of configuration data :
lsmod |grep -i nvidia
nvidia_uvm 2539520 0
nvidia_drm 151552 73
nvidia_modeset 2183168 33 nvidia_drm
nvidia 16334848 618 nvidia_uvm,nvidia_modeset
drm_ttm_helper 16384 2 nvidia_drm
video 81920 1 nvidia_modeset
et voici la BOM de ma nouvelle machine :
Système d’exploitation : AlmaLinux 10.1
Version de KDE Plasma : 6.4.5
Version de KDE Frameworks : 6.18.0
Version de Qt : 6.9.1
Version de noyau : 6.12.0-124.45.1.el10_1.x86_64 (64-bit)
Plate-forme graphique : Wayland
Processeurs : 32 × Intel® Core™ i9-14900F
Mémoire : 64 Gio de mémoire vive (62,4 Gio de RAM utilisable)
Processeur graphique : NVIDIA GeForce RTX 4070
Fabricant : XSM
Nom du produit : 760D5
Version du système : V1.0
Thank you for the proposal. Here is the answer from my system :
jacques@localhost:~$ sudo dnf install $(nvidia-detect)
bash: nvidia-detect: commande inconnue…
Additional information : when I start the session in “rescue” mode, I do not have the screen resolution problem…
Before running the nvidia-detect command, you need to set up the elrepo repository:
# rpm --import https://www.elrepo.org/RPM-GPG-KEY-v2-elrepo.org
# dnf install https://www.elrepo.org/elrepo-release-10.el10.elrepo.noarch.rpm
Thanks toracat,
I have install the repo, and the nvidia command told me I need the 580.142 driver which I tried to install, but the system told me this is in conflict with another version of driver already installed (595.58.03-1)…
Sorry, I forgot to tell you that you need to uninstall all nvidia-related packages before installing the elrepo driver.
OK. After removing previous NVIDIA files, and installing the $(nvidia-detect) output, it is still the latest driver version which is installed and does not solve the screen resolution.
So, enough for this time; I will reinstall AlmaLinux from scratch. thanks guys for your help !
Hello,
When you boot in secure mode the “nomodeset” parameter is loead instead the nvidia driver.
Try adding this to the kernel parameters
nvidia-drm.modeset=1
To do this, while in grub, press the ‘e’ key to edit the parameters:
Check if you have the parameters marked in green (you should have them) and then add the aforementioned parameter (marked in red)
Then, press Ctrl-X or F10 to boot.
Note:
If you format your system and attempt to install from RPMFusion using akmod-nvidia, first check if Secure Boot is enabled.
If it is, generate the necessary signatures on your system using kmodgenca before installing the module. This will ensure the module is signed when compiled.
Regards.
I solved the same problem 1 month ago on my serveur dell T340 with GPU 3080, when i update new kernel after : kernel-5.14.0-611.41.1.el9_7.x86_64 old nvidia libraries supersede new libraries icompatibles and packaged with new kernel ! this plan to solve my micmac:
sudo dnf remove “nvidia”
sudo rm -rf /lib/modules//weak-updates/nvidia
sudo rm -rf /usr/lib/modules//extra/nvidia
List and remove old incompatibles versions nvidia driver
sudo rm -f /usr/lib64/libnvidia-*.so.590.44.01
Rm old linked : sudo rm -f /usr/lib64/libnvidia-ml.so.1
Temporarly until almalinux nvidia new driver are ok :
sudo dnf update --disablerepo=almalinux-nvidia
sudo ln -s /usr/lib64/libnvidia-ml.so.580.142 /usr/lib64/libnvidia-ml.so.1
systèm compute new nvidi link TO new installed DNF versions
sudo ldconfig
If necessary : sudo dnf reinstall nvidia-x11-drv-libs --disablerepo=almalinux-nvidia
if screen instability : sudo ln -sf /usr/lib64/libGLX_nvidia.so.580.142 /usr/lib64/libGLX_nvidia.so.0
rebuild kernel
sudo depmod -a $(uname -r)
sudo dracut -f --verbose
My config now :
Driver ELRepo 580.142 (StabCUDA 13.2 (le) screen X11 + 120/240Hz (Fluide) alma repost no activate Démon Persistenced OK CUDA 13.2
reactive deamon kde here
verify: nvcc --version ; nvidia-smi
Driver and stability work better on 9.7 with X11 than Waylan !
sudo systemctl enable nvidia-persistenced
sudo systemctl start nvidia-persistenced
sudo ldconfig
Verify link are on driver 580 nvidia driver : ls -l /usr/lib64/libnvidia-cfg.so.1
Regards
Henri
Hello,
I do not think the AlmaLinux 9 workaround is directly applicable to AlmaLinux 10.1.
According to the official AlmaLinux documentation, the recommended approach is to enable the AlmaLinux NVIDIA driver repository and install the driver stack from that path, rather than mixing multiple sources or fixing libraries manually.
So I think it would be better to keep the NVIDIA-related repositories and packages consistent, instead of combining different repos or carrying over AlmaLinux 9-era workarounds into AlmaLinux 10.1.
Official documentation:
Regards
1 Like
With NVidia GPU’s the initial choice is between the open source drivers that are included in AlmaLinux (Nouveau?) and drivers from NVidia.
The drivers from NVidia do have two parts:
- kernel module
- the rest – the bits that Wayland and/or Xorg do use
The kernel module is either supplied as prebuilt, or is “assembled” in your machine by dkms or akmod.
The kernel module may also be signed, either out of the box or by the assembly tool.
Being signed is required, if Secure Boot is enabled (in UEFI) and certificate related to the signature must be in UEFI.
What AlmaLinux does provide is a prebuilt kernel module signed by same key as rest of AlmaLinux kernel, for which certificate was loaded to UEFI during install of AlmaLinux.
AlmaLinux does not provide “the rest”, but enables NVidia’s repository to get the rest from there.
One could install the kernel module from NVidia’s repository too, but then one has to set up the certificate (if Secure Boot is used).
NVidia’s repo does provide two versions of kernel module: “proprietary” and “open”.
The proprietary version supports GPU generations up to the 4000-series (like that 4070).
The open version supports newer GPU generations, starting with the 2000-series (Turing).
The other repos – for example ELRepo and RPM Fusion – do provide both kernel module and the rest, and handle Secure Boot in their way (if at all).
Simultaneous use of multiple repos that all do provide the same driver (NVidia, ELRepo, RPM Fusion) with incompatible packages will require exclusion of some packages.
AlmaLinux 10 has Wayland and no Xorg. The NVidia’s own drivers have had, and probably still continue to have some issues with Wayland. That is, with properly installed driver.
Try:
lspci -kd ::0300
The device type ::0300 is graphics. (Device type ::0200 is NIC, btw.)
Granted the -d ::0300 does list other than NVidia devices too.
1 Like