[525.78.01] NVIDIA card HDMI output freezes after launching fullscreen applications

I have this problem with fullscreened Shatterline launched via Proton, and fullscreen Minecraft both launched on NVIDIA GPU (offload) with AMD APU as the main card. Present only on Wayland.

So if I launch some applications with my NVIDIA GPU in fullscreen mode on an external HDMI display connected to my laptop, external HDMI output will freeze, but the application will not crash and even will keep rendering on NVIDIA GPU. If I try to replug the frozen display or just disable-enable it via KDE Plasma settings, my whole graphics system will freeze, and after that I’m not even able to switch to a different TTY.

Problem started to appear after a relatively recent NVIDIA driver update.

nvidia-bug-report.log.gz (632.8 KB)

Retested this bug with the older 470.161.03 NVIDIA driver version, and this bug is not present with the 470.161.03 driver version.

Here is nvidia-bug-report.log from 470.161.03 for comparison:
(470.161.03) nvidia-bug-report.log.gz (1.4 MB).

NVIDIA 525.78.01, bug is still here.
New bug-report:
nvidia-bug-report.log.gz (647.2 KB).

Maybe any debugging ideas to provide more details?

I have a similar issue also on Wayland and 525.78.01 (Manjaro & Gnome). The error occurs while playing full screen games on Proton. No errors on X11.

1 Like

A bit more details about my previous attempts to retest this bug.
Retest was done at the same laptop, the same OS version (snapshot) and even with the same KDE Plasma and Kwin versions. The only thing that I changed in my system is the official NVIDIA driver version (installed G05 instead of G06 [see the official NVIDIA driver repo for openSUSE Tumbleweed]).
Also this bug appeared during 5xx driver updates, but not during upgrade from 4xx to 5xx.

Here what my zypper will do if I ask him to downgrade to the G05:

root@yspace ~# LANG=C zypper install nvidia-computeG05
Loading repository data...
Reading installed packages...
Resolving package dependencies...

Problem: the installed nvidia-computeG06-525.78.01-16.1.x86_64 conflicts with 'nvidia-computeG05' provided by the to be installed nvidia-computeG05-470.161.03-58.2.x86_64
 Solution 1: Following actions will be done:
  deinstallation of nvidia-computeG06-525.78.01-16.1.x86_64
  deinstallation of nvidia-computeG06-32bit-525.78.01-16.1.x86_64
  deinstallation of x11-video-nvidiaG06-525.78.01-16.1.x86_64
  deinstallation of x11-video-nvidiaG06-32bit-525.78.01-16.1.x86_64
  deinstallation of cuda-runtime-12-0-12.0.0-1.x86_64
  deinstallation of cuda-runtime-11-8-11.8.0-1.x86_64
  deinstallation of cuda-12-0-12.0.0-1.x86_64
  deinstallation of cuda-11-8-11.8.0-1.x86_64
  downgrade of cuda-12.0.0-1.x86_64 to cuda-11.4.4-1.x86_64
  deinstallation of nvidia-gfxG06-kmp-default-525.78.01_k6.1.3_1-16.1.x86_64
  deinstallation of nvidia-glG06-525.78.01-16.1.x86_64
  deinstallation of nvidia-glG06-32bit-525.78.01-16.1.x86_64
 Solution 2: do not install nvidia-computeG05-470.161.03-58.2.x86_64

Choose from above solutions by number or cancel [1/2/c/d/?] (c): 1
Resolving dependencies...
Resolving package dependencies...

The following package is going to be downgraded:
  cuda

The following 5 recommended packages were automatically selected:
  nvidia-computeG05-32bit nvidia-glG05 nvidia-glG05-32bit x11-video-nvidiaG05 x11-video-nvidiaG05-32bit

The following 7 NEW packages are going to be installed:
  nvidia-computeG05 nvidia-computeG05-32bit nvidia-gfxG05-kmp-default nvidia-glG05 nvidia-glG05-32bit x11-video-nvidiaG05 x11-video-nvidiaG05-32bit

The following 11 packages are going to be REMOVED:
  cuda-11-8 cuda-12-0 cuda-runtime-11-8 cuda-runtime-12-0 nvidia-computeG06 nvidia-computeG06-32bit nvidia-gfxG06-kmp-default nvidia-glG06 nvidia-glG06-32bit x11-video-nvidiaG06 x11-video-nvidiaG06-32bit

1 package to downgrade, 7 new, 11 to remove.
Overall download size: 272.5 MiB. Already cached: 0 B. After the operation, 416.8 MiB will be freed.
Continue? [y/n/v/...? shows all options] (y):

NVIDIA 525.85.05.
Bug is still here.
For some reason appears less often in 2D apps/scenes.
New report file: nvidia-bug-report.log.gz (646.9 KB).

Just got this bug on X11.
On X11 I’m able to control the lagged external display colors via NVIDIA Xserver Settings, and just torn off-on the display in system settings in order to “fix the display”.
Also there was a system lag after turning my external display off.

NVIDIA, fix your crap, I beg you: nvidia-bug-report.log.gz (692.9 KB).

Wayland? A pinned post says not everything works on wayland EGL (egl is made for apps that want and older more stable GL, not for new features).

I don’t think nVidia owes Ubuntu KDE Suse more packges (infact i wish they’d help wider platforming not favor 2 or 3, personally). You didn’t say what OS your using what version, what version of Xorg (X11 is incorrect).

“wayland” is an “X wrapper” as is gtk or kde. it is continually changing versionally - and i wouldn’t be API are stable (ie, already released features canceled in new API - which KDE and GTK continually due against all rules).

to make matteres worse, Xorg has rapidly been hacking wayland features “that can’t be removed” from X as well as GL features that cann’t be removed

this makes predicting “what one means by saying on wayland” wildly un-predictable as to determining a cause.

Ok, seems like the bug report files do not include all the system info.
Here is some more info, hope this will be useful:
PCI.txt (10.8 KB)
XServer.txt (66.7 KB)
KWin.txt (7.4 KB)
Wayland.txt (29.4 KB)
Vulkan.txt (139.6 KB)
OpenGL (GLX).txt (82.4 KB)
OpenGL (EGL).txt (48.0 KB)
inxi -F.txt (3.2 KB)

maybe try phoronix for trolling instead of bug reports?

1 Like

I think I’ll just desolder that peace of garbage out of my laptop.

The problem is that no errors are logged apart from sometimes an Xid 56, “Display Engine error”, telling only what you already know, the nvidia gpu has an issue with its hdmi port.
Likely won’t help, but did you already check for a bios update?

Updated the BIOS, no effect.

Is there a way I can provide more info or logs for this report?

Here’s another user with the same issue:
https://forums.developer.nvidia.com/t/prime-usage-of-external-hdmi-monitor-results-in-xid-errors-with-dxvk-vkd3d/245370?u=generix
The common grounds so far are (X)wayland + amd igpu + nvidia offload. Though impossible to be sure to tell from only two datapoints.
I guess this bug was introduced with the change in

2022-11-28 version 525.60.11
    * Improved the performance of PRIME render-offloaded applications.

but wasn’t fixed with

======== Release 530 Entries ========
    * Fixed a bug that prevented PRIME render offload from working for
      Wayland applications when running on a system with an AMD iGPU.

Though that likely addressed a different bug.
At least the other person found a workaround by using gamescope.

1 Like

Don’t know when this happened, but seems to be fixed now (v535.86.05).
Here is the bug report file just in case: nvidia-bug-report.log.gz (906.5 KB)

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.