Hi Jeff. I delayed this upgrade as long as I could hoping the posted advice would help.
After doing the distro upgrade I find that I boot into nouveau.
I attempted the fix in your previous reply to kittyra’s similar post.
"I would strongly suggest that for f37 you do the following.
sudo dnf remove *nvidia* --noautoremove --exclude=nvidia-gpu-firmware
sudo dnf install akmod-nvidia-470xx --disablerepo rpmfusion-nonfree-nvidia-driver --enablerepo rpmfusion-nonfree
Wait 5 minutes then run dnf list installed *nvidia*
If that last command shows kmod-nvidia as installed then reboot and the nvidia card should now properly load the nvidia driver."
Unfortunately this has failed and I still reboot into nouveau.
My secure boot is off as always.
dnf list installed *nvidia*
Error: No matching Packages to list
I have the cuda-fedora35 and the cuda-fedora34 repos enabled. The cuda-fedora37 does not install properly. Could my problems have something to do with the repos?
I appreciate all you help with past issues of this sort.
The nvidia drivers installed from cuda-fedora and those from rpmfusion seem to conflict.
Please try these simple stepa first and see what the results are.
sudo dnf remove kmod-nvidia-* and when that completes run sudo akmods and wait at least 5 minutes for that to complete properly.
After the second command completes run dnf list installed kmod-nvidia*. This should now show only the kmod-nvidia package for the matching running kernel.
If it does so then now reboot and tell us the final result.
Note that having the cuda-fedora repos enabled for older releases and having any packages from those repos installed may interfere with the nvidia packages from rpmfusion. You showed the cuda package from rpmfusion is installed so you should not need any thing from the cuda-fedora repos.
To verify if you do or do not have anything from the cuda-fedora repos please run dnf list installed | grep -i cuda-fedora and post the output here please.
Also, to see what is installed that is nvidia related the command should be dnf list installed '*nvidia*'. The one you showed with ‘NVIDIA’ in all caps really is not indicative of a clean nvidia install.
There probably was a package in the directory where that command was run that had nvidia in the name. Surround the *nvidia* with single quotes like '*nvidia*'to avoid shell globbing from interfering.
Thanks Jeff you nailed it. There were 6 fedora35 kmods still running. Your procedure sudo dnf remove kmod-nvidia-* and sudo akmods fixed the nouveau boot problem.
I understood that I required the cuda-fedora packages to build CERN’s root application to utilized the gpu . Not certain that it is still needed. However it is clearly messed up on my machine.
There were some problems with dependencies upgrading with the cuda-fedora37 repo enabled.
Probably I should remove most leaving only a consistent cuda version.
You have a lot of packages installed from cuda-fedora34 and cuda-fedora35. It seems possible that there may be problems related to the age of those packages when running fedora 37. Note that most things cuda related use the GPU for computing and since the cuda packages have not been updated while the gpu drivers have been, there are potential problems.
I am glad that at least the update of the drivers to the rpmfusion version fixed the GPU problem for you.