Seems people have issues with Nvidia 545 drivers lately

Sometimes we struggle with Nvidia drivers… With the latest 545 series, their first “stable” tagged version had some regressions: Linux, Solaris, and FreeBSD driver 545.29.02 [new feature branch release] - Announcements and News - NVIDIA Developer Forums

Their current production ready driver is Linux, Solaris, and FreeBSD driver 535.129.03 [production branch release] - Announcements and News - NVIDIA Developer Forums

With 545.29.06 most of the regressions of 545 series got fixed. We pushed that one to our stable branch as there were no issues reported in testing and unstable branches regarding that driver. However, it seems now they pop up in stable branch.

So let us know if you have any and also check your dmesg and lsmod plus modprobe nvidia before posting …

3 Likes

I see only a few has problems in stable:
No issue, everything went smoothly 92% shows the pool vote.
And 3% who had a issue was able to fix it.

The flickering nvidia driver was a bigger issue, that we had for 6 month, since april.

Anyways, i feel the pain for this few that run into a issue now.

i have trouble on sceen suspend , login appears , and black screen after ( cursor mouse is ok )
i have to restart lightdm.service

i don’t notice any difference with the new driver.

i had a weird glitch in steam were i was getting a black screen when starting a game,
so i switched from proton 8.0.4 to experimental and back again.
also suddenly there was a “Cyberpunk” update and now all is as usual.

i started get this error/message but apparently that’s not an issue:
kernel: NVRM: loading NVIDIA UNIX x86_64 Kernel Module 545.29.06

see here: NVRM: loading NVIDIA UNIX x86_64 Kernel Module 545.29.02 Thu Oct 26 21:21:38 UTC 2023 - Linux - NVIDIA Developer Forums

The forum has many issues reported against 545 driver series. If you have one read also this before posting about it: »»»»»»»»»» If you have a problem, PLEASE read this first «««««««««« - Linux - NVIDIA Developer Forums

I still have my particular flickering issue on the newest nvidia 545 build from the recent update.

Changing user and screen locking is broken on XFCE and possibly KDE Plasma,
and some testing seems to point to the new drivers

(see:
Xfce4 black screen with cursor when switching between users
After 2023-12-01 I can not get in to a timed out session
Unable to login on resume due to the “empty” login screen
)

Also,since Nvidia recommends 535 as the current production/stable driver here:
https://forums.developer.nvidia.com/t/current-graphics-driver-releases/28500
I believe it would be great for manjaro to allow the user to select the driver version that works best for them, or to perhaps downgrade to 535 until the last bugs have been ironed out.

Also, is there anyone waiting for better stability that is waiting before updating and therefore not replying to the pool vote?

1 Like

I have an issue on Wayland only

CS2 Flickers

All games run through Proton run better on Wayland however.

On x11 CS2 runs as it should.

For me nvidia modules somehow fails to load at kernel 6.6.* so i had to switch back to 6.1.69-1.

jounalctl -b:

Dez 26 17:42:46 Manjaro-Legion5 kernel: nvidia: loading out-of-tree module taints kernel.
Dez 26 17:42:46 Manjaro-Legion5 kernel: nvidia: module license 'NVIDIA' taints kernel.
Dez 26 17:42:46 Manjaro-Legion5 kernel: Disabling lock debugging due to kernel taint
Dez 26 17:42:46 Manjaro-Legion5 kernel: nvidia: module verification failed: signature and/or required key missing - tainting kernel
Dez 26 17:42:46 Manjaro-Legion5 kernel: nvidia: module license taints kernel.
Dez 26 17:42:46 Manjaro-Legion5 kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 235
Dez 26 17:42:46 Manjaro-Legion5 kernel: 
Dez 26 17:42:46 Manjaro-Legion5 kernel: nvidia 0000:01:00.0: enabling device (0000 -> 0003)
Dez 26 17:42:46 Manjaro-Legion5 kernel: nvidia 0000:01:00.0: vgaarb: VGA decodes changed: olddecodes=io+mem,decodes=none:owns=none
Dez 26 17:42:46 Manjaro-Legion5 kernel: NVRM: loading NVIDIA UNIX x86_64 Kernel Module  545.29.06  Thu Nov 16 01:59:08 UTC 2023
Dez 26 17:42:46 Manjaro-Legion5 kernel: nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  545.29.06  Thu Nov 16 01:47:29 UTC 2023
Dez 26 17:42:46 Manjaro-Legion5 kernel: [drm] [nvidia-drm] [GPU ID 0x00000100] Loading driver
Dez 26 17:42:46 Manjaro-Legion5 kernel: ACPI Warning: \_SB.NPCF._DSM: Argument #4 type mismatch - Found [Buffer], ACPI requires [Package] (20230628/nsarguments-61)
Dez 26 17:42:46 Manjaro-Legion5 kernel: ACPI Warning: \_SB.PC00.PEG1.PEGP._DSM: Argument #4 type mismatch - Found [Buffer], ACPI requires [Package] (20230628/nsarguments-61)
Dez 26 17:42:46 Manjaro-Legion5 kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x24:0x72:1436)
Dez 26 17:42:46 Manjaro-Legion5 kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
Dez 26 17:42:46 Manjaro-Legion5 kernel: [drm:nv_drm_load [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to allocate NvKmsKapiDevice
Dez 26 17:42:46 Manjaro-Legion5 kernel: [drm:nv_drm_probe_devices [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to register device

Well, we had for each Nvidia driver a package and the user had to select which driver he wanted. It also meant that a user had to update the driver and was more or less “stuck” on a driver series if not doing that. Each series got updated until Nvidia dropped a series.

Then most users wanted the rolling Nvidia driver back, which had some benefits but also some regressions. Drivers in general is always something …

1 Like

I am also experiencing various issues related to version 545.
My window flickers during all glmark2 tests except the first one ([build] use-vbo=false).
I have encountered very similar problems in Minecraft, making the game unplayable. Enabling vertical synchronization in the game slightly improves the situation, but does not fix it completely due to small inconsistencies between fps and monitor frequency. Enabling/disabling VBO in the game settings does not fix the problem.

I am also experiencing problems in Unigine Superposition, with black lines and a flickering screen that seems like sync issues. The same problem occurs in the Sea Of Thieves game, even in the main menu.

I have been trying to roll back the driver version to 535 for the last ~15 days, but have been unsuccessful. Unfortunately, I did not notice the problems in the games right away, and the timeshift snapshot has already been overwritten by other updates.

I attempted to use the following command, but my GNOME session does not start with it:
sudo manjaro-downgrade --cached-only linux66-nvidia nvidia-settings lib32-nvidia-utils nvidia-utils mhwd-nvidia libxnvctrl
Adding linux66 and linux66-headers to the above command and/or removing the drivers in mhwd before the command also does not help.

I understand that I could possibly roll back the version using nvidia-535xx-dkms from AUR, but I am afraid of the compatibility issues with Manjaro and potential problems with future updates.

As a result, I have reverted back to version 545, because version 470 in mhwd adds other problems (GNOME animations don’t work).

Could you please add an option to select version 535 in mhwd or provide a working method to roll back the version? Thank you.

1 Like

I have had similar problems in various games for a month now, since nvidia 545 driver came out.
Mostly flickering and black lines on screen too. I haven’t found a proper way to rollback to nvidia-535xx and I’m hoping to find a solution for this lasting and uncomfortable situation.

Having a option to select previous drivers versions would be great, and prevent users to be stuck with glitchy drivers in the future too.