NVRM: GPU at PCI:0000:01:00: GPU-1dfdd8a3-609f-ab69-f862-db0ee121b8ce
NVRM: GPU Board Serial Number: 0
NVRM: Xid (PCI:0000:01:00): 79, pid='<unknown>', name=<unknown>, GPU has fallen off the bus.
NVRM: GPU 0000:01:00.0: GPU has fallen off the bus.
NVRM: GPU 0000:01:00.0: GPU serial number is 0.
NVRM: Error in service of callback
RTX 3060 laptop, can not get nvidia-bug-report after that, script just hangs. In fact any process that tries to use nvidia GPU after that error just hangs and can not be killed even with SIGKILL.
NVidia guys, what I want to conclude you rushed up with final release. System detailed were given in posts above, but only new in KDE 6.1.1. Problem is with last beta 555.52.04 all fine for me, with 555.58 KDE’s taskbar and whole plasma shell simply hangs up an stops responding in some half an hour. So final relase in definitely defective and last beta was better, then final release. I believe you rushed up making quick fines without testing and marking it stable.
Mistakes happen, so just letting you know and hoping for any time soon fixes coming. So far staying with 555.52.04, as with Plasma shell hanging that often is no go for daily use.
Also found out what causes it to hang. this thing supposed to show window preview from buffer. And it always hangs with it with picture from buffer after which stops reacting. Hand symptom not like loop, where cpu/gpu busy to 100%, but opposite like infinite i/o wait, where only option remains is kill -9 plasmashell_process_id (without -9 flag no reaction). So something messed up. Community says no such issues with other GPUs, only for nvidia it happens. After killing plasmashell as you can see picture from buffer no longer being drawn and no hangs anymore. So it’s somehow that thing related. Got this thing even on beta driver, so all 555.XX series affected. Happens this when mouse near task icons drawing this and swithing windows around with ALT+tab
I have horrible random flickering and random black screens after updating to 555 with both Wayland and X when using 4k and anything above 60Hz. This did not happen with 550.
RTX 3080
Fedora 40 with GNOME 46
Driver version 555
Kernel 6.9.5
I had to set NVreg_EnableGpuFirmware=0 to get a smoother experience. Without setting it, switchng workspaces feels chonky. Apps flickering is gone on wayland, but firefox was continuously crashing until I set MOZ_WAYLAND_ENABLED=0 and this caused the performance to drop drastically. youtube has some lags and stutters as well. Blender and Unreal engine are crashing after updating to 555. This issue was non existent on 550.
There’s something strange going on with GPU power states compared to 550 driver: if an application uses nvidia GPU but currently does not run any operations on it (pauses when window is minimized for example), the GPU can go to idle state, which causes application to freeze momentarily when it tries to resume operations (GPU needs time to wake up). And sometimes GPU falls off the bus (Xid 79) when this happens. I disabled GPU firmware to see if makes any difference, for now no Xid errors, I will test more.
Sorry for just asking without going through everything, but i can not resize any window in latest wayland kde (endeavourOS) with nvidia 555 stable without flickering and crashing the app.
It happens to kde-native apps, gtk apps, electron apps, all of them.
Resizing will lead to flickering, freeze and crash within a few seconds.
I’m using a multimonitor setup with 1x 8k and 4x 4k, maybe i should try single-monitor first, wow i have the best ideas.
Update: I disabled everything but 1 4k screen and no more flickering or crashes so far.
Update2: The big bad 8K TV seems to be the issue, as soon as i activate it the problems return.
bug :nvidia-drivers-555.58 DVI output not working ->black screen
looks like this not an isolated case and seems to affect various cards.
my case:
using a NVIDIA GeForce GTX 1050
kernel 6.9.7
using card with curent set up no problem for 5 years
noticed has I have a 2 monitor set-up (same monitors)
one use HDMI the other DVI
booting display correctly dmesg on card DVI output to monitor up to when when nvidia takes over with X/wayland/DE monitor conected via dvi then screen turn black
As I see it, there are some stuff that still needs urgent attention.
Multimonitor VRR support under wayland.
The crashes in unreal editor/blender
The VRAM handling (like Xwayland VRAM keeps growing until it reaches a certain limit, no real shared VRAM etc)
The XID 109 crashes (the last of us and other titles)
+1 on this and also the newly introduced Flip Event Timeout on Serious Editor(Serious Sam modding tools) this wasn’t an issue prior to adding fbdev, but even if fbdev is set to 0 it would still trigger it after opening editor - sometimes immediately upon opening, other times while resizing UI.
Holds true for 555.58.02. I can reproduce this with Knossos: start the application with render offload, minimize it and wait monitoring the state of the GPU. It will go to D3Cold after several seconds. Now restore application’s window and you’ll get visible freeze until GPU gets to D0 again. At least with firmware disabled it doesn’t fall off the bus.
A quick follow up to this:
Updating my drivers to 555.58.02 has fixed my crashing issues (on CS2)
Before updating, the game would crash with or without GSP enabled, but now I can play full matches without any VRAM related issues with GSP enabled.
My roommates machine can no longer resume from suspend with 555.58. MSI B550 Tomahawk, Ryzen 5700G, XFCE (should be kernel 6.9.7-arch1-1). Waking the machine results in a kernel panic style screen talking about nVidia IRQ issues and reboot needed and it’s locked up hard (cue “No reisub for you!” joke).
I had it die a different time where the FB was better resolution so more was on screen, which is when I saw the nvidia IRQ issues bit with needs reboot but there was no camera to snap a pic around. The above screenshot is what came after that from earlier testing.
A quick follow up to this:
Updating my drivers to 555.58.02 has fixed my crashing issues (on CS2)
Before updating, the game would crash with or without GSP enabled, but now I can play full matches without any VRAM related issues with GSP enabled.
I’ve noticed 555.58.02 has made it take longer before I get a Xwayland crash, but it still crashes as soon as my 3080 hits 9.8GiB/10GiB used. Can you try filling your VRAM and checking if it still crashes? I’ve done it by opening an intensive game and a youtube video on the background.