(Sorry for my language)
For some reason, my gnome shell started to slow down on the last 2 kernels. You can’t select Wayland in GDM. (5.14.0-570.32.1.el9_6.x86_64)
I installed drivers from both elrepo and rpm-fusion.
If you remove the drivers, wayland starts and gnome shell doesn’t slow down.
I don’t quite understand, but it used to work (though I can’t say for sure whether it was wayland or x11), but it definitely didn’t slow down with real nvidia drivers.
I copy what @tqhoang mentioned somewhere else.
I did find that using the NVIDIA Settings app and setting the PowerMizer from Auto/Normal → Performance mode kind of resolves it. There are 5 power levels/limits (0-4). Probably running in anything higher than 0 would resolve it.
May be related?
Thank you.
I understand what you mean.
That gdm script file with conditions that each may disable Wayland? (E.g. “driver 510 or older”)
Some of the conditions use power-related NVidia utilities that at least earlier (550?) ELRepo packages did not provide, but NVidia’s own packages (at least from 570) did. That (lack of utility) too does disable Wayland. I was actually surprised when I switched to NVidia’s packages (for other reasons) and was suddenly in a Wayland session.
@sergmx - I am in the process of upgrading the ELRepo NVIDIA driver to the latest production 580.76.05. I’m working to make Wayland work out of the box too.
I have it 0-2 (Quadro P2000) and yes it works! Thanks.
/usr/lib/udev/rules.d/61-gdm.rules
Thank you very much.
Sorry, I was in a hurry, now I worked longer and the slowdowns appeared again. Although I have the maximum GPU frequency enabled.
I don’t remember this on older kernels. Although maybe I had wayland enabled there.
Older than: 5.14.0-570.32.1.el9_6.x86_64, 5.14.0-570.30.1.el9_6.x86_64
Indeed. It has about 12 points, where it uses GOTO="gdm_disable_wayland"
At most two of them are due to the version of the driver.
At least five of the conditions (that can lead to gdm_disable_wayland) do require certain files/services.
RPM Fusion xorg-x11-drv-nvidia-power and NVidia’s nvidia-driver packages do provide those files/services.
ELRepo’s 570 packages do not. @tqhoang wrote above that the next version will be better.
Rocky does use Nouveau driver with NVidia GPU, if there are no “third-party” drivers installed.
The Nouveau supports Wayland but cannot use all features of the GPU.
Sorry, but the driver installation instructions say that you just need to do: akmod-nvidia.
And it doesn’t handle any xorg-x11-drv-nvidia-power.
I didn’t know that you also need to install this.
Thanks.
FYI - the ELRepo drivers for NVIDIA 580.76.05 on RHEL 9.6 are available now.
dnf install kmod-nvidia --allowerasing
Reboot and then make sure to select Wayland on the GDM login screen.
Note:
This thread (topic) has been moved as it does not relate to the site or the forums.
Thank you very much.
I removed all 46 unnecessary packages that rpm-fusion pulled. I installed new drivers from el-repo and wayland also worked!
P.S. For some reason, the egl-wayland package that was previously installed from el-repo was not installed.
I don’t see egl-wayland in ELRepo mirrors.
Rocky’s appstream has 1.1.13.1-1.el9
NVidia’s cuda-rhel9-x86_64 has 1.1.20-1.el9
Glad to hear it is working for you.
Regarding the egl-wayland dependency, we now bundle the newer lib version that NVIDIA includes…same for egl-gbm and egl-x11 (for EL10).
Basically we added conflicts/provides to the nvidia-x11-drv RPMs.
True. If one looks with dnf rq -l egl-wayland.x86_64
one sees that the package has practically one file in it.
Then one can look for the files with
dnf provides *lib64/libnvidia-egl-wayland.so.1\*
and should see that nvidia-x11-drv-libs-580.76.05-1.el9_6.elrepo.x86_64 has them too.
Whether to build multiple small packages or couple big ones is up to whatever is more convenient / leads to least amount of hassle.
Sorry, I got confused with the drivers.
A new driver has been released: 580.82.07.
I decided to check what’s going on in elrepo and I don’t understand anything.
I have 580.76.05 and I still don’t see them on elrepo.
How come?
It does now look like:
# dnf -q --enablerepo=elrepo,elrepo-testing list --showduplicates kmod-nvidia kmod-nvidia-open kmod-nvidia-470xx
Available Packages
kmod-nvidia.x86_64 570.172.08-1.el9_6.elrepo elrepo
kmod-nvidia.x86_64 570.172.08-2.el9_6.elrepo elrepo
kmod-nvidia.x86_64 570.181-1.el9_6.elrepo elrepo-testing
kmod-nvidia.x86_64 570.181-1.1.el9_6.elrepo elrepo-testing
kmod-nvidia-470xx.x86_64 470.256.02-1.el9_4.elrepo elrepo-testing
kmod-nvidia-470xx.x86_64 470.256.02-2.el9_5.elrepo elrepo-testing
kmod-nvidia-470xx.x86_64 470.256.02-3.el9_6.elrepo elrepo-testing
kmod-nvidia-470xx.x86_64 470.256.02-4.el9_6.elrepo elrepo-testing
kmod-nvidia-470xx.x86_64 470.256.02-5.el9_6.elrepo elrepo-testing
kmod-nvidia-470xx.x86_64 470.256.02-5.1.el9_6.elrepo elrepo-testing
kmod-nvidia-open.x86_64 570.172.08-1.el9_6.elrepo elrepo
kmod-nvidia-open.x86_64 570.172.08-2.el9_6.elrepo elrepo
kmod-nvidia-open.x86_64 570.181-1.el9_6.elrepo elrepo-testing
kmod-nvidia-open.x86_64 570.181-1.1.el9_6.elrepo elrepo-testing
While elrepo-testing for el10 has:
kmod-nvidia-580.76.05-1.el10_0.elrepo.x86_64
kmod-nvidia-open-580.76.05-1.el10_0.elrepo
Excuse me, but why did you remove 580.76.05 for rl9?
And some 570.181 appeared, why is that?
rpm -qa | grep nvidia
nvidia-x11-drv-libs-580.76.05-1.el9_6.elrepo.x86_64
kmod-nvidia-580.76.05-2.el9_6.elrepo.x86_64
nvidia-x11-drv-580.76.05-1.el9_6.elrepo.x86_64