Very weird indeed – and very frustrating. It’s kinda crazy having a notebook but being unable to let it suspend. Also, what’s even more puzzling is that automatic suspend fails 100% of the times, while manual suspend works 99% of the time (I’ve seen it fail a couple of times as well, but very few times). So, it’s not that it can’t suspend, it’s something that is done when it is triggered by idle time (as it should be) that makes it fail. To avoid having to constantly hard-reboot, I disabled suspend when charging, and try to use the notebook plugged to the charger when I’m working.
I’m way out of my league here, but seeing errors like NVRM: dmaAllocMapping_GM107: can't alloc VA space for mapping. made me wonder if this could be somehow related to the fact that I am using nvidia NVreg_PreserveVideoMemoryAllocations=1 ? Do you have this enabled as well?
Just noticed that with the 570 driver installed, initramfs size increases significantly, on the order of 50 Mb. Hard to say if all of that can be attributed to the driver. Some may be kernel driver includes. For boot partitions of smaller size, this might create an issue if including the driver in the boot img.
I am on CachyOS with a AMD 7900X, RTX4070ti Super, and two displays. Have tried both the proprietary and open kernel modules.
Upon resuming from sleep, dmesg shows the following error: [drm:__nv_drm_semsurf_wait_fence_work_cb [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to register auto-value-update on pre-wait value for sync FD semaphore surface
Immediately upon resuming, the desktop environment (KDE on Wayland) is unresponsive for quite a while. Icons cannot be pressed, the clock occasionally becomes corrupted.
I have also experienced problems launching games after the machine has been to sleep. Splash screens will display, then turn black once the game has launched. Changing window focus to a window on display 2 will cause the game to become visible, until you change focus back to the game, where it goes back to a black screen. nvidia-bug-report.log.gz (1.4 MB)
Anyone else seeing some insane blur with NVIDIA-SMI 570.86.16 ? Specifically in CP2077 benchmark scene. The blur is unplayable. This is using “auto” adaptive sync in kde plasma. FG on, Transformer model set to auto/quality. Cachyos with proton-cachyos (native) , also tried proton (bleeding edge), no difference. Seems to be very hit and miss though. Sometimes launching the game it will be fine and clear. Then if I exit and relaunch (still full screen mode) it will be blurry the next time.
Toggling between fullscreen and windowed borderless sometimes fixes it, but its completely random. Sometimes I start the game and it works fine, other times not at all. For now I need to disable framegen. It is unpredictable.
I regret to report that setting fbdev=0 does NOT fix the problem (can’t really say why my previous test worked). Just double-checking if I did it right, here’s my nvidia.conf file:
I regenerated all initramfs by running dracut --regenerate-all --force, and rebooted. If the charger is unplugged, internal display stays dark all the way, only option is to hard-shutdown by holding the power button.
Unfortunately my display is not working for me at all when VRR is engaged on HDM2.1 with a single display. All I get is a black screen. I can alt tab and disable and then it works fine. Bug report attached.
Getting black screen for 1-2 seconds whenever I alt-tab out of a VRR enabled application on my 4k 240hz display. This is the same issue that was in a previous Windows driver. This behaviour is not on present Windows driver. But is certainly here on the current Linux driver.
Missing you meant? I haven’t verified in a while but I recall the .run file installs that to /etc/vulkan/icd.d instead, you may want to check there too.
Some distros do prefer the system path rather than the one intended for users’ own files though (aka, Gentoo does /usr/share/vulkan for one, either path works).
On fedora 41, kernel-6.12.11-200 and nvidia-open-570.86.15 the nvidia module fails to load. Dkms succesuflly builds the modules but at boot they won’t load and modinfo finds only the version for the previous kernel-6.12.10-200.
Same issue with VRR over HDMI 2.1 - I have a G80SD 4k 240Hz monitor that works fine until VRR is engaged. Once engaged the screen goes black and no signal can be found. This does not happen over DisplayPort but this monitor has issues with DP that force me to use the HDMI connections. This can easily be replicated by setting KDE’s Adaptive Sync to “Always” or simply firing up any fullscreen VRR engaged app. 4090 on Gentoo here - works perfectly with the 565 drivers (which I have rolled back to).
