You are not logged in.
Pages: 1
Hello,
Currently I use nvidia 387.22-3, but this was happening quite often with previous versions of NVIDIA proprietary drivers. I have GTX 1080 Ti GPU, maybe that can be the cause, since when I had GTX 780 I had no this issue. Now I use KDE 5.39.0 / Plasma 5.11.2 with kwin_x11 utilizing OpenGL 3.1 desktop effects.
The problem is that graphical user interfaces freezes from time to time (before mostly I was having Graphics Reset and not freeze), mostly it happens when I launch Google Chrome or Steam, but sometimes it happens randomly. Every time it happens I see such text in dmesg:
[14095.376345] NVRM: GPU at PCI:0000:01:00: GPU-5eea2366-8f3b-df5c-910b-2c2f8e023d3d
[14095.376347] NVRM: GPU Board Serial Number: 0320717****48
[14095.376348] NVRM: Xid (PCI:0000:01:00): 31, Ch 00000020, engmask 00000101, intr 10000000
Then I simply get the Xorg to unfreeze by killing kwin_x11 process from tty2 (sometimes plasmashell too) and relaunching them. Maybe someone knows a way to avoid this problem?
Can it be a hardware problem and not software?
Offline
Now it got even worse... Every launch of Google Chrome makes this to happen. Not sure what to do. If I will not find any solution, might put back my GTX 780 card...
Offline
http://docs.nvidia.com/deploy/xid-errors/index.html
Memory page fault, driver or HW issue.
Do you have a sufficient PSU?
Offline
http://docs.nvidia.com/deploy/xid-errors/index.html
Memory page fault, driver or HW issue.
Do you have a sufficient PSU?
Oh, thanks for reply. I have 750W PSU (digital) and I see max usage with high GPU load is 250-300W, normally when I do stuffs it's around 100-120W. I don't think that PSU can cause problems. About driver also not sure, I wonder if other GTX 1080 Ti owners has any issues. If the hw is faulty, then I need to send card back...
Offline
This is usually™ a driver bug.
Try to disable the compositor. If that prevents the error showing up, try to force an older GL API (2.0)
In addition, you could try to (significantly, at least to 381.xx) downgrade the driver and see whether that makes a difference. You'll have to bypass pacman for this, though, what is a bit nasty. (Your GPU is only supported with 367.27 and up)
Offline
Thank you for the tips! I will try an yes, when compositor is disabled, this does not happen, just sometimes when I launch app which utilizes OpenGL or GPU in general (Steam Games)...
Offline
Just the same error here with an Asus GTX 1060 6GB. It is a couple of driver versions that I got this, all freezing, have to restart sddm, pretty annoying because sometimes it happen also with virtualbox. Best triggers: chrome, VB.
Offline
For me best triggers are also like you mentioned Google Chrome and few others including Steam. Was unable to solve the problem and sent my GTX 1080 Ti to warranty service... I took my old GTX 780 and put to computer. Everything seems smooth first one hour and then same problem. I think it's faulty driver in general. Instead of Google Chrome, I tried out Vivaldi (seems to be based on Chromium if I'm correct) and this problem does not occur with browser (yet), just now with Steam. But I don't use it everyday, just when I'm in mood for some gaming
berny99, my solution for this is, switch to tty2, login as root, kill kwin_x11, logout, go to tty1, launch "kwin_x11 --replace" (using krunner) it's annoying to do this, but no need to restart sddm.
Offline
adomas, I resolved like this, i'm using kde. Hope it might be useflu to you also:
Disabled triple buffering on xorg settings, disabled tearing prevention (never) in kde settings monitor settings.
Enabled force full composition pipeline in nvidia settings, it's one week without crash, finger crossed.
As a side effect I have also no tearing at all on the screen.
Offline
oh, cool, I will try this thank you berny99!
By the way warranty service said that my GTX 1080 Ti is "impossible to fix" and gave me a refund but as with GTX 780 I have same issue, I will try your solution.
Offline
Notice that this will draw more GPU power. Seems like buffer swap related.
Is it reproducible w/o making use of old buffers?
export KWIN_USE_BUFFER_AGE=0
kwin_x11 --replace &
Offline
This KWIN_USE_BUFFER_AGE=0 did not solve the problem. I still get xid 31:
[19534.909861] NVRM: Xid (PCI:0000:01:00): 31, Ch 00000058, engmask 00000101, intr 10000000
Now I'm pretty sure it's faulty nvidia driver... On both GPUs: GTX 1080 Ti and GTX 780 same issue.
Offline
I'm no help on your specific problem, but I run KDE with a GTX 1070 (it's a laptop, but a discrete card -- no Optimus junk) and I do not and have not had problems with any nVidia drivers. I don't run testing, and am current with updates. Is it possible that your issue is hardware related? Motherboard related?
Matt
"It is very difficult to educate the educated."
Offline
Few days ago I switched to Manjaro from Arch, and this issue happens way less often. Now once a day, not like every few hours...
Here in Manjaro forum one user has same issue:
https://forum.manjaro.org/t/after-2017- … ze/35205/3
As Xid error code is 31, it does not mean faulty hardware:
Xid 31: GPU memory page fault (Driver Error)
Offline
Actually the matrix says hardware or driver and the text includes the option of broken client code.
Anyway and wrt. to your manjaro situation:
Assuming you're using the same kernel and nvidia versions: https://wiki.archlinux.org/index.php/Mi … de_updates ?
Offline
For one month I was using Manjaro Linux and had these crashes once in a 2-3 days. I did not like some things about Manjaro, so I switched back to Arch Linux and now I have these crashes daily. It happens almost every time as I launch Chromium or Steam. Can it be related to KWin package on Arch? How did Manjaro made it happen not so often?
Is there a way to get more debug info than just dmesg output? This Xid 31 is killing me
Offline
Offline
Pages: 1