Ever since late Aug 2025, I’ve started getting nvidia driver crashes - CTX SWITCH TIMEOUT + NV_ERR_GPU_IN_FULLCHIP_RESET.
Across several nvidia driver versions, and full system upgrades. I’ve waited to see if new versions of the driver would fix it, but they do not.
This crash happens mainly when running KDE Plasma + Wayland. This has forced me to revert to using X11.
The crash still happens on X11, just way less often - however I’m also getting full screen freezes on X11 from which the system recovers most of the time. (1 second or so freezes)
I can make this bug happen on Plasma + Wayland basically under an hour, most of the time or at least once a day.
The issue seems to happen when i’m playing a video in Chromium based browsers (for example youtube), scroll down to the comments or tab out and read something in a different tab while the video still plays.
I’ve tried disabling hardware acceleration and it seemingly still happens. HW accelerated video decode is not enabled on nvidia + chromium anyways.
When the issue happens, the screen usually fully freezes while the mouse remains not frozen, then a few seconds later the mouse freezes as well, then further seconds later Plasma crashes fully and a black screen appears. The only way to recover is to ssh into the pc and issue a reboot - nothing else works to recover the system softly.
I’ve only tried the nvidia-open drivers, since i cannot run the closed drivers with a 50 series card.
I’m attaching a nvidia-bug-report.
This issue is making the computer barely usable. I cannot use Wayland at all and the full screen freezes on X11 make using the computer a chore. Gaming on Windows, the gpu does not seem to crash, so I don’t think it is a hardware issue.
To eliminate potential issues with my specific configuration, I have installed a new fresh OS install on a spare SSD and the issue still happens. This has to either be an issue with the firmware, drivers or the card itself.
Full system details:
Operating System: Arch Linux
KDE Plasma Version: 6.5.5
KDE Frameworks Version: 6.22.0
Qt Version: 6.10.2
Kernel Version: 6.18.7-arch1-1 (64-bit)
Graphics Platform: X11 (now since i can’t use wayland)
Processors: 16 × AMD Ryzen 7 9800X3D 8-Core Processor
Memory: 32 GiB of RAM (30,4 GiB usable)
Graphics Processor 1: NVIDIA GeForce RTX 5070 Ti/PCIe/SSE2
Graphics Processor 2: NVIDIA GeForce RTX 5070 Ti/PCIe/SSE2
Manufacturer: Micro-Star International Co., Ltd.
Product Name: MS-7E59
System Version: 2.0
@amrits Thanks. Sadly the issue doesn’t have an “easy“ reproduction case.
It can take 1 hour to reproduce, it can take 1-2 days. It happens fairly randomly when regularily using the computer. When using Wayland, it happens at least once a day for me.
Good luck trying to reproduce it and hopefully you can find something in the error report.
Today, the gpu drivers once again randomly crashed in X11.
The previously attached bug report is from wayland, where it crashes around once a day.
Today’s crash is in X11 (only ever had 2 of these crashes).
The full situation of the crash was:
Watching youtube video in a chromium browser (brave), video ended and without me touching the computer a minute or two later the screen went black, and the computer had to be rebooted over ssh.
@amrits This has been going on for months now, and I haven’t managed to figure out a reliable repro scenario. As far as I can tell, it just happens randomly. Mostly noticeable with a video playing and scrolling in some other tab, but just generally happens during regular use, sometimes when not even interacting with the computer. In X11 the random 1s freezes are every couple minutes and very noticable.
One thing I forgot to mention in the first post is, that the issue started more or less 2 weeks since the last system update I’ve made at that time. (first time it happened was aug 28th 2025, last update before that was the 15th). So it started happening effectively without a driver/kernel update, and since the issue appears at least once a day or two in Wayland, 2 weeks until it first happens is odd if it were caused by the update.
This does open the possibility of some firmware corruption/hardware issue. However besides the random hiccups in X11 and full freezes in Wayland, the GPU seems to work correctly, even gaming in Windows.
Is there some recommended way to fully test the GPU hardware to eliminate the issue being there?
Since it happened in a fresh Arch installation, I would say the issue is either driver or hardware related.
Thanks for sharing your experience and the thread in Fedora forums.
Going through the thread, it really seems like the exact same issue. There are some CTX_SWITCH_TIMEOUT/GPU_IN_FULLCHIP_RESET threads here, but mainly with specific identified game settings and not random when watching youtube or just existing in chromium.
Can you try confirming the second part of my post as well? That is, try switching KDE to an X11 session and see if you get random 1s screen freezes that the system recovers from most of the times. (one freeze per maybe 15-30 minutes, more than one per hour for sure) Mainly noticeable when watching videos.
In my case it at least makes the system halfway usable, since in Wayland, it usually leads to a full gpu hang and having to fully restart the system.
If that gets confirmed, we can hopefully get more bugreport archives here get some attention on this matter… I’ve been living with this for 6 months now, across many kernel and nvidia driver releases with no improvement in sight.
I’ve upgraded to the nvidia-open version 595.58.03, and the issue still happens.
I might be using the computer a bit differently, it feels like it’s crashing a bit less, but i still got 2 crashes since updating to 595 maybe 2 weeks ago.
@amrits I’ve updated to that version around 3-4 days ago and so far no crash.
Since it really mainly crashes on youtube, I’m also trying the hardware video decoding hypothesis once again. I’ve removed the community nvidia libva driver maybe 2 weeks ago and no crashes since then as well. However given the random nature of the crashes, I can’t guarantee it won’t crash soon.
Since the community libva-nvidia-driver gets installed by default on Arch when using the archinstall script, could you try including that in your experiments? It possibly could be triggering some weird driver hang.
What is the configuration of the system you’re testing it on?
I’m having the same exact issue. I made an account just to add on. I’m running Fedora 44 KDE Plasma (6.19.14-300.fc44.x86_64) and Nvidia (595.58.03 from akmod-nvidia/RPMfusion) on Wayland with an RTX 5070. I’ve been having random freezes. It’s usually per app. For example, I will try to open a tab in firefox and it freezes for 1-15 seconds, or typing in Discord, typing in Konsole, or even just opening a folder in Dolphin. Most of the time it recovers, but every so often the entire computer will freeze. I think it’s mostly centered around apps that are using hardware acceleration?
Update: I switched to akmod-nvidia-open and it has the same issue. One thing I’ve thought of, could it be related to a multi-screen setup? I’m running three screens and will try using one for awhile.
I’m running on a single 1440@165hz monitor. I don’t think high refresh rate has effect, but at least in my case multi-monitor is not a concern.
@akmod-nvidia-open vs akmod-nvidia, not too sure about how the RPMfusion drivers work, but guessing both of them load the same open kernel modules, since the 50 series requires them and won’t work with the closed drivers.
I’ve updated to 595.71.05 and the issue still persists.
I’m not sure if this is related, but I see these errors when firefox freezes:
May 07 07:30:16 pony org.mozilla.firefox[3491]: Error writing selection data: Error writing to file descriptor: Broken pipe
May 07 07:30:16 pony org.mozilla.firefox[3491]: Error writing selection data: Error writing to file descriptor: Broken pipe
May 07 07:30:16 pony org.mozilla.firefox[3491]: Error writing selection data: Error writing to file descriptor: Broken pipe
May 07 07:30:33 pony plasmashell[2541]: QWaylandDataOffer: timeout reading from pipe
May 07 07:30:33 pony plasmashell[2541]: QWaylandDataOffer: error reading data for mimeType text/plain;charset=utf-8
May 07 07:30:34 pony plasmashell[2541]: QWaylandDataOffer: timeout reading from pipe
May 07 07:30:34 pony plasmashell[2541]: QWaylandDataOffer: error reading data for mimeType text/plain
Here is what I see when Konsole freezes:
May 07 07:45:00 pony plasmashell[2541]: QWaylandDataOffer: timeout reading from pipe
May 07 07:45:00 pony plasmashell[2541]: QWaylandDataOffer: error reading data for mimeType text/plain;charset=utf-8
May 07 07:45:01 pony plasmashell[2541]: QWaylandDataOffer: timeout reading from pipe
May 07 07:45:01 pony plasmashell[2541]: QWaylandDataOffer: error reading data for mimeType text/plain
May 07 07:45:12 pony kwin_wayland[2408]: Failed to find a working output layer configuration! Enabled layers:
May 07 07:45:12 pony kwin_wayland[2408]: src KWin::RectF(0,0 3440x1440) → dst KWin::Rect(0,0 3440x1440)
May 07 07:45:12 pony kwin_wayland[2408]: drmModeListLessees() failed: Permission denied
May 07 07:45:12 pony kwin_wayland[2408]: Atomic modeset test failed! Permission denied
May 07 07:45:12 pony kwin_wayland[2408]: Applying output configuration failed!
May 07 07:45:12 pony kwin_wayland[2408]: Failed to find a working output layer configuration! Enabled layers:
May 07 07:45:12 pony kwin_wayland[2408]: src KWin::RectF(0,0 1920x1080) → dst KWin::Rect(0,0 1920x1080)
May 07 07:45:12 pony kwin_wayland[2408]: Failed to find a working output layer configuration! Enabled layers:
May 07 07:45:12 pony kwin_wayland[2408]: src KWin::RectF(0,0 1920x1080) → dst KWin::Rect(0,0 1920x1080)
May 07 07:45:12 pony kwin_wayland[2408]: Failed to find a working output layer configuration! Enabled layers:
May 07 07:45:12 pony kwin_wayland[2408]: src KWin::RectF(0,0 3440x1440) → dst KWin::Rect(0,0 3440x1440)
@slackwar I think your issue could be different to mine.
My issue is specifically CTX SWITCH TIMEOUT and NV_ERR_GPU_IN_FULLCHIP_RESET spammed in journal after a full driver crash leading to a browser and kwin crash. It’s a full screen freeze leading to a black screen only recoverable with a full system reboot.
You can check out my logs in the bug report archives and compare. If they look different, it might be best to open a separate forum post.
I have done two things and the issue has not happened since:
Updated to 595.71.05
Added pcie_aspm=off to my kernel on boot
I will keep running like this for a while, and if it stays stable, remove the kernel boot argument for further testing. Fingers crossed. Also thanks to @amrits for the official support!