Cannot successfully install drivers using the documentation and recommended tool on ubuntu server 22.04.Official guide: https://ubuntu.com/server/docs/nvidia-drivers-installation
Run sudo ubuntu-drivers install --gpgpu nvidia:535-server
and sudo apt install nvidia-utils-535-server
. After reboot run nvidia-smi
and get this:
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.
When run modprobe nvidia
getting:
modprobe: FATAL: Module nvidia not found in directory /lib/modules/5.15.0-106-generic
When run lspci -k | grep -A 2 -E "(VGA|3D)"
, getting:
2a:00.0 VGA compatible controller: NVIDIA Corporation AD102 [GeForce RTX 4090] (rev a1) Subsystem: NVIDIA Corporation Device 167c Kernel modules: nvidiafb, nouveau
Using other ways like local runfile work fine. So why the recommended way does not work?
UPDATE: could solve it, there are 2 distinct problems. 1st is that, it appears, drivers with -server
suffix do not work properly with consumer-grade GPUs like Nvidia 4090. 2nd is that there is nouveau
in Kernel modules
even though I explicitly blacklisted it. For some reasons, blacklisting does not work strictly if kernel module is already built. Also, running sudo ubuntu-drivers install --gpgpu nvidia:535
does not force kernel module rebuilt. What worked is purge nvidia files from the system sudo apt-get purge nvidia*
, sudo apt autoremove
. Then run sudo ubuntu-drivers autoinstall
and reboot. Here the autoinstall
did the trick, trying to directly specify a particular version with nvidia:535
fallback on nouveau
in my case. After this, it is possible to re-run sudo ubuntu-drivers install --gpgpu nvidia:535
with a particular version specified and everything works.