I had issues getting cuda to work on a debian distro.
I’ve tried several driver versions but seem to have messed something up. I can succesfully run
sudo apt-get remove --purge nvidia-*
but when trying to install the latest driver I get this message:
root@Server:~# aptitude install nvidia-driver-430
...
The following packages have unmet dependencies:
nvidia-kernel-common-430 : Conflicts: nvidia-kernel-common but 20151021+4 is to be installed
xserver-xorg-video-nvidia-430 : Depends: xserver-xorg-core (>= 2:1.19.6-1ubuntu2~) but 2:1.19.2-1+deb9u5 is to be installed
nvidia-installer-cleanup : Conflicts: nvidia-driver-binary which is a virtual package, provided by:
- xserver-xorg-video-nvidia-430 (430.14-0ubuntu0~gpu19.10.1), but 430.14-0ubuntu0~gpu19.10.1 is to be installed
- xserver-xorg-video-nvidia-418 (418.74-0ubuntu1), but it is not going to be installed
The following actions will resolve these dependencies:
Keep the following packages at their current version:
1) nvidia-dkms-430 [Not Installed]
2) nvidia-driver-430 [Not Installed]
3) nvidia-kernel-common-430 [Not Installed]
4) xserver-xorg-video-nvidia-430 [Not Installed]
Any idea what I can try to make it work? Reinstalling the OS is not an option.
I installed driver 390.116 now via aptitude install nvidia-driver and as opposed to previously (I’ve tried this driver before) I can see the devices with nvidia-smi.
Is there a way that I can install a more recent driver so I can use CUDA 10.x?
Your answer helped me put 2 and 2 together, I learned that there are ‘testing’ and experimental packages available for debian, cuda 10.0.130 is currently experimental and I got it installed.