Weird, I did some more testing in gnome, and when I came back to plasma I stopped having the issue. While I’m glad I’ve been able to launch some games and at leat benchmark the 7700s, if I don’t know what fixed it I don’t know why/how it was broken before, and it may break again at any point.
Edit: To little surprise, the issue returned with little indication as to why.
OK I just made a small test here with 6.12, 6.11 and 6.10.
They screwed up something between kernel 6.11, if my memory is correct, something between 6.11.3 and 6.11.6. Using 6.11.9 it freezes as well.
If you want to do gaming on Plasma 6, I would recommend you to downgrade to kernel 6.10 latest (6.10.9 ? ), it works fine to me
I’ve been having this issue for now. Is there a bug report filed somewhere so that people can know about this kernel regression? Would be good to get it fixed soon.
As a side note, I’ve found running games through gamescope appears to fix things as well, if you’d rather do that instead of a kernel downgrade.
Nope!
It’s pointless to report a bug without any useful information. I have no idea what causes it, I just know it happens.
Also, not sure if this is truly a bug introduced by AMD kernel team, or Kwin uses an obsolete/bad way to access GPU, or some issue specific from Framework itself (BIOS/hardware)
Using kernel 6.12.1 also gives me random artifacts at screen. Not always, but happens.
We could try and do a kernel bisection to figure out the exact change that causes the issue. I’d say that’d be useful enough information for a bug report. It’s not a process I’m all too familiar with, though, but I’ve seen it done before to pinpoint similar issues.
It seems to me this problem starts where this problem ends.
I don’t think it was properly fixed, or it caused another problem.
That update broke Ultimaker Cura, saying “This program requires OpenGL 2.0 or higher.”
As per a guide i found about the same issue on windows, I’ve deleted all of its files i could find, aside from the appimage itself, but it has no effect.
With this new problem, I used to mainly get temporary screen freezes whenever a fullscreen game has literally anything displaying over it. it got worse over time with consistent permanent freezes, but setting the display refresh rate to match the game and disabling adaptive sync made it only temporary again.
I still get freezes on vlc too, although more rare than ever now.
Worst thing is that nothing shows up in crashed process viewer this time.
If you’re getting content not updating when launching (certain) games fullscreen, but can sometimes bring up the start menu (which makes the game display again), you might be enjoying this particular bug: [BUG] [PSA] : GPU bug causes crashes on KDE Plasma.
For some folks the KWIN_DRM_NO_DIRECT_SCANOUT workaround seems to be fine, for me it doesn’t help at all, and when I play games that struggle like this I’m switching to swaywm for now (or you could use Gnome) until the bug is resolved.
I’ll give that environment variable a try later. I have my doubts about it being the same issue, though, as that seems to have been a problem for much longer than this has. As others already mentioned, this issue didn’t exist before kernel 6.11, and it looks like that issue has been around for months (maybe years?). 6.11 was released before that post was made, but I don’t think most distros were on it yet… iirc fedora wasn’t, at the very least. But someone else in the thread mentioned the issue being around for 2 years. Could be related, but I’m unsure.
I wasted away my weekend grinding out some raider.io in WoW, and learned a couple things about this issue I’m encountering. It appears to be quite inconsistent. I was able to launch a game fine twice in a row, but then on the 3rd launch some time later it failed at the frozen black screen. A reboot allowed me to successfully launch once more, but it appears to me as if whether a launch will be successful or not is entirely random, and if it fails, it will always fail until something changes (e.g. via a reboot).
I am more than happy to provide logs for a bug report, but I do not know what logs to grab.
I had something similar a while back, and in my case, it was the interposer.
I had to remove it, clean the contacts with a lens-cleaning pump, and install them back.
Alright it’s been a while but I finally tested the environment variable change and it had no effect, issue is still there. So seems to be a different issue as I suspected. Also updated bios and drivers today because that update appeared and it did nothing too. Still thinking it’s a kernel issue. Unfortunately I can’t really downgrade to 6.10.9 because that version was never built for fedora 41.
Could be a Wayland issue?
appears to be. Just tried X11 and that works fine. So we can add that as a workaround.
It is a Wayland issue, but not sure if KWIN is responsible. I think something changed from AMDGPU and broke KWIN in the process. The issue only happens after 6.11.3 or 6.11.4.
But I do prefer wayland because X makes my life miserable with VFIO GPU. It crashes because it doesn’t disconnect properly as Wayland does.
I think it’s kinda related with HVM but not sure
Installed some old kernels and found 6.11.3 works fine but 6.11.4 does not, so hopefully that narrows things down a little.
I know we don’t have a ton of info - but I do really think we should report this regression somewhere. Seems like a pretty major issue to me and kernel regressions are generally taken quite seriously.
edit: looking into doing a kernel bisect now to narrow down the issue further, if I can find an exact commit that caused the issue I think that would be more than enough to report.
Is this only happening to you with the dGPU? I’m having similar issues but also with the iGPU. I’m using Fedora 41, so I’m also on a very new kernel and tried it today on Ubuntu 24.04, which worked fine. I only noticed this issue after upgrading to the 03.05 firmware. Maybe it is also related to that. What firmware are you on?
I was having the issue before the 3.05 firmware/bios update as well, so there’s that. I’ll see if I can run the game off the integrated graphics later - currently in the midst of the kernel bisection which takes a while.
Ok! Completed the bisection and got a lot of new info here. First off, the offending commit is drm/amd/display: Re-enable panel replay feature. Interestingly, it looks like there was already some problems it caused on FW16 that have been discussed, some very similar to this issue - however, from what I can find in bug reports, the issue was supposed to be fixed in 6.12, which it obviously has not fully been. Last bug report that’s related is here, seems to be the same issue.
I’ll build latest myself just to make sure - already tried it from fedora rawhide but those kernels aren’t always the same as from the source tree so I’ll make sure, if still broken there I’ll reopen the issue over on freedesktop.
Ok, so now I’m really confused. Compiled 6.13 myself and it works fine. Fedora’s rawhide of the same version (even same rc!) does not. And given that OP was on arch and using 6.12 (the version where it was meant to be solved) it seems something is different with both Fedora and Arch’s kernels from the source tree. Going to look into this some more, if anyone has any ideas let me know.
Edit: just confirmed that the fix has still been committed to fedora’s kernel, so I really have no clue why it’s still broken.
I’m running Manjaro Plasma unstable here with kernel 6.12.4-1-MANJARO currently, and I ran into this with my Framework 16 iGPU (I don’t yet have the dGPU). I couldn’t use adaptive sync at all, or the display would freeze constantly. I read the bug linked here, and I tried what this post suggested and added amdgpu.dcdebugmask=0x400 to my kernel parameters. Now I’m able to use adaptive sync in display settings, as well as play Trackmania in full screen mode.