I’ve experienced frequent freezes as well, and switching from Ubuntu 24.04 to CachyOS doesn’t fix the issue either. Both the display and the backgroudn process will start to lag then just freeze completely while the background tasks go back to normal per the sound output. In fact, the freeze seems to occur more frequently with CachyOS.
I disabled pcie aspm and that might be why Ubuntu lasts longer but it will still freeze:
We are currently debugging the issue based on shared logs due to lack of local reproduction.
Would really appreciate if any one can try repro by setting the nvidia-settings → Powermizer → Preferred mode - set to Prefer Max Perf and share the test results.
I’ve configured Powermizer to “Prefer Max Perf” and the behavior remains unchanged.
However I’m not sure how it should be behaving.
When I first enable powermizer, it stays in the maximum performance level and at 2400 MHz, but it does not stay there forever and still drops down to lowered clock speeds. Is that how “Max Perf” is supposed to behave?
Unrelated to Powermizer, but related to the issue.
Every time the issue happens, rather than resetting the computer with the power button, I ssh in and perform a regular reboot.
Ever since the error log changed to “Received signal from GSP that PMU has halted.”, the computer reaches the phase where it should halt and reboot, but it gets stuck there - either on a black screen or showing the last log line - the part where it should halt. I need to then reset the computer with the reset button - it won’t actually reboot on it’s own.
I am having the same issue, now it happens almost daily if not more then once a day
It seems to always happen when I am playing back a video either Twitch or Youtube (although it might just be cause I almost always have one playing, but it never happened without one of the two in the background)
Hardware tests:
I did a full PC cleanup
Run MemTest86+ with 2 passes and 0 errors
GPU power draw/clocks/temps all normal during playback (30-46W, 40-43°C)
PSU voltages/load also normal
More changes I tried:
I disabled NVreg_EnableGpuFirmware
I removed the libva-nvidia-driver since it was causing issues not loading for Firefox, still did not change (although it lasted for a full day, almost 2, without freezing?)
I removed some plasmoids causing potential playback issues (i dont know if it mattered, i was just kinda doing any change at that point)
vm.compaction_proactiveness is set at 20, no changes
I am currently testing with pcie_aspm=off but i doubt it will change much…
EDIT: forgot to mention that I noticed a kwin_wayland segfault that occurs at every login/session start, with a stack trace showing threads stuck in libnvidia-eglcore.so, the session auto restarts normally after and I have no idea if this is anything useful or not, and I also have no clue how I would go about solving it if it was
Specs:
OS: Fedora 44 KDE
KDE Plasma Version: 6.7.0
Kernel Version: 7.0.12-201.fc44.x86_64
Always using Wayland (since x11 is disabled on Fedora 44)
GPU: NVIDIA GeForce RTX 4070 Ti
NVIDIA Driver: tested on 595.80, 595.71.05, and currently 595.58.03
I tried switching to CachyOS as well but that only seems to be doing worse. At least for Ubuntu 24.04 it only happens only a couple of times per day, while for CachyOS it seems to be every half hour or so. I’ve also disabled the iGPU and the bluetooth stack but that doesn’t seem to help either. However, disabling the bluetooth stack seems to fix another freeze issue (that is worse since it also blocks sysreq) which isn’t relevant here.
More specifically, it seems like it is causing xid 32 problems all the time from reading the journal log. I can be more specific as well, not only does the mouse not hang, the audio does not hang either. Then after a couple seconds both would hang. Unfortunately for me, gnome can’t recover from it, so I have to force a reboot with sysreq every time and can’t generate a log file containing the error. The following journal log might be relevant:
Jun 28 15:31:18 marshland kernel: NVRM: GPU at PCI:0000:01:00
Jun 28 15:31:18 marshland kernel: NVRM: Xid (PCI:0000:01:00): 32, pid=5199, name=gnome-shell, channel 0x00000007
Jun 28 15:31:18 marshland gnome-shell[5199]: Window manager warning: META_CURRENT_TIME used to choose focus window
Jun 28 15:31:20 marshland Xorg: Logitech G502 HERO Gaming Mouse: SYN_DROPPED event - some input events have been lost.
Jun 28 15:31:22 marshland Xorg: event processing lagging behind by 28ms, your system is too slow
Jun 28 15:31:25 marshland Xorg: event processing lagging behind by 48ms, your system is too slow
Jun 28 15:31:25 marshland Xorg: NVIDIA: Wait for channel idle timed out.
Jun 28 15:31:26 marshland polkitd: Unregistered Authentication Agent
Jun 28 15:31:26 marshland gnome-session: Failed to acquire idle monitor proxy: Message recipient disconnected from message bus without replying
Jun 28 15:31:26 marshland gsd-power: Failed to acquire idle monitor proxy: Message recipient disconnected from message bus without replying
Hello everyone. In a desperate move, I’ve temporarily swapped GPUs with a friend.
It’s a 4070 Ti, so a generation older than my 5070 Ti.
It’s been almost a week with me constantly trying out scenarios, that were basically guaranteed to crash the gpu on a system running for a week. And well, it doesn’t crash. It’s been rock solid ever since I booted it up. The system went through at least 10 sleep/wake cycles as well.
I haven’t changed anything regarding the system configuration to eliminate software changes (no updates of any kind), running kernel 7.0.14 with nvidia-open 610.43.02.
This opens up 2 possibilities for me:
The GPU has a hardware defect.
There have been changes in GSP between 40 and 50 series making the 50 series crash.
I’m more closely leaning towards number 1, mainly because of other threads talking about “PMU has halted” and “CTX SWITCH TIMEOUT”, which are not always on the 50 series. I’ve seen multiple pieces on 40, 30 or even 20 series cards.
@amrits How do you see this? Should I RMA the card?
Well, the 4070 Ti card was fully stable, no issues.
Swapped it back and borrowed another friend’s 5070 Ti (different manufacturer), and it crashed within 4 hours of being installed. Literally the same crash as with my card.
@amrits Any updates on your side? I can’t return the card now, RMA seems not worth it since it seems to be 50 the whole series having issues (or at least 5070 Ti), and the card is literally worthless to me now, since I can’t keep it running more than a day without it crashing.
Do you have any actionable steps? Is there some internal priority in getting this fixed? How can the GPU not last a day without crashing, it’s insanity.
Literally the only thing I can do is run on my CPU integrated gpu and hope you guys fix it, or sell the card and DOWNgrade to 40 series (or buy from competition).
@amrits Thanks for keeping me updated. Shame you still cannot reproduce the issue.
You already have my hardware information, so I’ll just add onto that.
Whenever I can, I’m upgrading my BIOS version (currently 7E59v2AC3 on the x870e tomahawk), so at the very least if it’s something AMD/AMI/MSI are aware of, it’s not fixed in the latest version (if the BIOS/mobo is a contributing factor)
Since I think I never mentioned it, my specific GPU is GAINWARD GeForce RTX 5070 Ti Phoenix V1 16G, but since another model from a different OEM behaved the same it’s probably not the GPU itself. Not sure what model was the 4070 Ti not exhibiting the issue.
Would you be able to share what the differences are between the 4070 Ti and 5070 Ti in terms of GSP firmware? I think one of the big tells could be the fact at least my exact issue isn’t happening on the 4070 Ti.
Is there anything else I can do to help? If you guys can’t reproduce it, there might be something else I can do to help getting it resolved.
At this point, I’m not really sure what to do otherwise. The GPU is now unplugged and I’m forced to run my daily computer usage off the integrated GPU, and I’m considering if there’s a way to return it within warranty period (would have to be a long argument with the shop, probably not successful) or sell it second hand at a loss and get something else. I would love to avoid doing it, but the incredible instability is not really a daily driver material for me.
After a few weeks of running on the iGPU, i tried plugging in the GPU again.
The issue still happens on 610.57.04.
I’ve also tried limiting the PCIe speed to gen 4 for the gpu slot to eliminate signaling degradation at gen 5 speeds. This did not help and the crash still occured.
I’ve been having the NV_ERR_GPU_IN_FULLCHIP_RESET error in multiple of my journalctl logs after having my system suddenly freeze (prompting me to do a reboot with the magic key combination: “REISUB”).
I’m currently using CachyOS, but i also had freezing issues on my X11 fedora 43 system, which started around the 27th ofJuly - after i updated my system.
System specifications:
Motherboard: (MS-B928)/MS-B9181
CPU: Intel Core i5-9400F
GPU: GeForce RTX 2060 SUPER Ventus OC 8GB - MSI
Driver: nvidia v: 610.57.04 alternate: nouveau,nvidia_drm
non-free: 550-6xx.xx+ status: current (as of 2026-07) arch: Turing
Kernel: 7.2.0-1
DE(WM): Niri - Wayland
I’m not sure if the nvidia bug report script ran successfully, but i’ll attach it.
I also have the same issue with black screens and effectively locking up the computer since I’m forced to reboot as another tty also doesn’t come up. This seems to happen most when I’m playing Civ5 and multiple times a day depending on what I’m doing. I do not have to be gaming to see this issue.
Mon Sep 28 16:41:39 2026
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 615.71.09 KMD Version: 615.71.09 CUDA UMD Version: 13.4 |
+-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GeForce RTX 4070 Off | 00000000:01:00.0 On | N/A |
| 0% 35C P8 14W / 200W | 1291MiB / 12282MiB | 9% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 1203 C+G /usr/bin/kwin_wayland 70MiB |
| 0 N/A N/A 1315 G /usr/bin/Xwayland 3MiB |
| 0 N/A N/A 1362 G /usr/bin/plasmashell 145MiB |
| 0 N/A N/A 1818 G ...share/Steam/ubuntu12_32/steam 5MiB |
| 0 N/A N/A 2022 G ./steamwebhelper 52MiB |
| 0 N/A N/A 2047 G ...am/ubuntu12_64/steamwebhelper 120MiB |
| 0 N/A N/A 2749 G /usr/bin/krunner 11MiB |
| 0 N/A N/A 2801 G /usr/lib/firefox/firefox 252MiB |
| 0 N/A N/A 3700 G /usr/bin/ghostty 200MiB |
| 0 N/A N/A 20754 G /usr/bin/systemsettings 70MiB |
+-----------------------------------------------------------------------------------------+
@amrits Any updates? It has been 2 months again since last proper response from nvidia.
The GPU is basically unusable paperweight for me. It would really be helpful if you guys could keep the GPU from crashing or at least regularily kept us updated.