However when I run dnf install akmod-nvidia I get a strange error:
# dnf install akmod-nvidia
Last metadata expiration check: 0:01:24 ago on Fri 19 Aug 2022 15:28:56.
Problem: conflicting requests
- package akmod-nvidia-3:510.68.02-1.fc36.x86_64 requires nvidia-kmod-common >= 3:510.68.02, but none of the providers can be installed
- package akmod-nvidia-3:515.65.01-1.fc36.x86_64 requires nvidia-kmod-common >= 3:515.65.01, but none of the providers can be installed
- package xorg-x11-drv-nvidia-3:510.68.02-1.fc36.x86_64 is filtered out by modular filtering
- package xorg-x11-drv-nvidia-3:515.65.01-1.fc36.x86_64 is filtered out by modular filtering
(try to add '--skip-broken' to skip uninstallable packages)
I’ve never seen this error and searching around didn’t bring anything useful… Any ideas?
akmod-nvidia-3:510.68.02-1.fc36.x86_64 and xorg-x11-drv-nvidia-3:510.68.02-1.fc36.x86_64 were updated to version 515.65.01-2 on rpmfusion-nonfree-updates.
Refreshing your package database should remove all references to 510.68.02-1 and eliminate the conflict. More aggressive alternatives if refresh doesn’t stop the conflict: sudo dnf clean dbcache
Removes cache files generated from the repository metadata.
This forces DNF to regenerate the cache files the next time it is
run. sudo dnf clean expire-cache
Marks the repository metadata expired. DNF will re-validate the
cache for each repository the next time it is used. sudo dnf clean metadata
Removes repository metadata. Those are the files which DNF uses
to determine the remote availability of packages. Using this op‐
tion will make DNF download all the metadata the next time it is
Simply go to NVidia website, download and install appropriate diver for your display card. Supported OS => Linux x64 and ARM, Solaris, FreeBSD, … any problem is linked to Red Hat/CentOS/Fedora approach to FOSS … For advanced frameworks you need Fedora 35.
This 3 step process always removes any older packages and installs the latest version for you thus eliminating the version conflict seen above. It also avoids the need to manually install and upgrade the modules as would be needed if you chose to follow the suggestion of @marko23 above…
Back when it was new, it gave quite a bit of update problems when modular packages conclicted with regular packages and things like that. It seems there are not many packages in fedora-modular any more.
WHat the original poster needs or has from fedora-modular I have no idea.
cuda-fedora33 is (was) a repository provided by nvidia company, and of course meant for Fedora33 . So maybe a couple of years ago that repository has been enabled and the package nvidia-driver from that repository is a modular package. Therefore nvidia-driver is still enabled as a module until you actually disable it. And that even long time after cuda-fedora33 has been removed as a repository.
If you want to check for your self here is the repository configuration
Then run dnf module list and you will find the mystery module.
Now, verious xorg-x11-drv-nvidia-xxx packages from rpmfusion also provides nvidia-driver and that is why you can’t install any of then as long as nvidia-driver is enabled as a module. The original poster actually did disable it and thereby solved the problem.
Please do not suggest that users use the binary blob from the Nvidia website. It causes all sorts of issues. All users should be please pointed to the RPM Fusion Nvidia drivers only—they are packaged up to work with the Fedora package set, and issues there should be debugged as we’re doing here.