Only if the drivers were still in the testing repository. Since they have been pushed to stable and all repos are synced, dnf downgrade will load the driver from rpmfusion-nonfree
akmod-nvidia.x86_64 3:570.133.07-1.fc42
2 Likes
Maybe you should consider testing the new drivers before they are pushed to stable.
^^’ I specifically switched to akmod-nvidia for dnf-automatic and to be able to … not test anything, since I’ve used the nvidia drivers from its .run format for years (and I mean it), but recently got (very) tired of it. Thank you for rpmfusion-packager, I will look into it.
I was able to retrieve all the necessary packages to switch back to 570.153.02, and everything is back to normal.
Now, understanding why 575.57.08 is not working for me is an entire another story that needs investigation (not sure if it’s appropriate to do it here).
1 Like
GPU: NVIDIA GTX 950M
OS: Fedora 42
DE: GNOME 48.2
Kernel: 6.14.11-300.fc42.x86_64
The 3:575.57.08-2.fc42 drivers installed just fine.
Wayland works relatively ok, but I can’t login to Xorg as it crashes while trying to start the x server and falls back to the login page.
Proprietary NVIDIA Drivers 3:575.57.08-2.fc42 kernel 6.14.9-300.fc42.x86_64:
- GNOME on Xorg (X11) → won’t login, the server x crashes and you’re back to the login page
- GNOME on Wayland → works just fine and everything is snappy as usual
Nouveau Drivers for the NVIDIA GPU, kernel 6.14.9-300.fc42.x86_64
- GNOME on Xorg (X11) → works but it takes a very long time for the various program windows to open up. Once they’re up, their response time is ok-ish but still lagging a bit.
- GNOME on Wayland → works but it takes a very long time for the various program windows to open up. Once they’re up, their response time is ok-ish but still lagging a bit.
The Nouveau drivers for the GTX 950M don’t allow for reclocking and therefore the GPU stays at the slowest possible clock at all times, hence the “sluggishness”.
575.64 is in the testing repo
1 Like
Thanks for the new driver, Scott.
Unfortunately, 575.64 didn’t work for my GTX 950M. 
GPU: NVIDIA GTX 950M
OS: Fedora 42
DE: GNOME 48.2
Kernel: 6.14.11-300.fc42.x86_64
Hi,
I installed it on Fedora 42 and it works fine with Nvidia 5070 Ti Laptop.
Thanks
Edit: Spoke too soon. There are problems with Chrome and KDE. Maximising window is buggy now! The CUDA is broken too.
Reverting back.
GPU : nvidia rtx 2060 FE
OS : Fedora 41
DE : xfce 4.18 / kde plasmashell 6.3.5 (wayland)
Kernel : 6.14.9-200.fc41.x86_64
Everything works great under xfce. I can’t find any obvious issue (but I didn’t search all that well).
kde is sluggish as hell (100% CPU), especially after running a konsole, but I’m unsure this has anything to do with the driver. I’m only reporting it as an observation, compared to the times I used the nvidia driver in its .run format.
This gave me headaches because I use an eGPU (GTX 1070 / Pascal), and I happened to update via dnf while the eGPU was disconnected. I realized this late because I only use the eGPU when booting a virtual machine (qemu), and the message errors I was getting were not clear at all (see this bug report that I opened and closed after finding the root cause "DMAR: DRHD: handling fault status reg 2" with vfio on kernel 6.13.11-200.fc41.x86_64, works with 6.13.9-200.fc41.x86_64 (#2953) · Issues · QEMU / QEMU · GitLab).
To be honest I don’t understand this change: it’s normal to have different lines/variants of a driver, and they usually are packaged under different package names. This runtime detection only adds complexity and hidden variables, in my opinion.
In light of the recent deprecation of the driver for a number of older generations of NVIDIA cards (https://www.phoronix.com/news/NVIDIA-580-Linux-Driver-Last-HW), how do you plan to provide the now legacy branch of drivers @leigh123linux ? It seems to me that the mentioned chip generations do overlap with the GSP-only cards line, so it would make sense to create an akmod-nvidia-closed package (akin to akmod-nvidia-open).
I doubt many people will run into the issue you describe.
The runtime detection makes the installation process easier for most.
akmod-nvidia-open isn’t easy to install and causes conflicts and other problems, that why it’s in the tainted repo.
In an ideal world we would use boolean requires to solve this.
Requires: (pkgA or pkgB)
Using the ‘or’ statement in any rpmfusion rpm package causes the repo creation tool (mash) to fail.
The 580xx driver packages are likely to continue to use runtime detection .
1 Like
@leigh123linux The latest open module driver sometimes fail with error:
nvidia 0000:01:00.0: can't suspend (nv_pmops_runtime_suspend [nvidia] returned -5)
All the desktop softwares (like system monitor MissionCenter/Resources, steam etc…) can’t detect the GPU exists except lspci, and system poweroff/reboot time will be much longer, and if you switch to other VT like press ctrl+alt+F4 the whole system will freeze. This bug appears after updated driver from 570.xx to 575.xx.
@leigh123linux I’m using RTX3060 laptop with AMD R7-5800H, do you know where is the problem?
I had another problem with the 575.64 driver installation from the rpmfusion updates repo. Running fedora 42.
Installation of the 6.15.4 kernel and the driver occurred at the same time.
However, the kmod-nvidia module for that kernel was not automatically installed.
I ran akmods to attempt to install it and akmods reported OK for the build, but still did not install the driver.
In order to install the driver after booting to the 6.15.4 kernel I had to manually run dnf to install the package that was left as /var/cache/akmods/nvidia/kmod-nvidia-6.15.4-200.x84_64.rpm
When I did that install it also installed the related kmod-nvidia-open package.
Note that when I did the update I was running the 6.14.9 kernel and it did properly install the updated driver for the running kernel, just not for the 6.15.4 kernel.