[TRACKING] Graphical corruption in Fedora 39 (AMD 3.03 BIOS)

Just chiming in with my own experience of this. See it as per others on resume, almost vanilla fedora gnome 39 install / 3.03 etc / 64gb ram / Linux 6.5.11-300.fc39.x86_64 x86_64. I only got half way through the thread because I’ve got to get to work shortly, but will loop back with more detail this evening if I find it hasn’t been solved (there were some promising things in posts a few weeks ago).

In my case it was just on battery. Would have been resuming Firefox, Thunderbird, Alacritty, Terminal, Docker with Postgres and Redis and a Rust Web server app. Of those Alacritty is gpu accelerated (and I kind of love the idea that the gpu intense task that triggered the graphical issues was a terminal…).

Not sure if it was suspended or hibernating, probably the former as far as I understand the defaults? Most of the screen was artifacting (White with some visible regions of the login down the right) so I couldn’t get in and just had to hard-reboot. The only different thing I’ve done with the machine is that yesterday I did a full reinstall from KDE Spin to Gnome because of cascading instability with KDE (though I suspect that was all software and my fault as I had been playing with tiling plugins without cleaning them up).

Will return later on and read the rest of the thread / provide more context / logs as necessary.

Updated w/Additional Info: Running 150% scaled, I haven’t attached an external monitor at all to the laptop yet (though I do have the HDMI expansion card in there), have only heard the fans once for a 30 second period in the week I’ve had it. Also was running Slack which I just spotted reported that it had crashed.

I too don’t think this looks like a hardware fault at this stage, definitely feels different to hardware failures causing graphical corruptions I’ve seen in the past. Onward to Kernel 6.6 :).

@mikeymop ; hate to break it to you but I’m running all the various included amdgpu patches that were added both for the 6.6 and 6.7 (6.7 has a lot more of them). And I am still getting freezes.

6.7 with SG disabled it’s far less frequent tho.

1 Like

Just chiming in with a datapoint: I’ve had my machine on Fedora 39 (XFCE/Xorg) for about two weeks now and have experienced no issues. I have not added any amgpu boot params. One thing I wonder is if this is related to the amount of memory in the system; I have 32G in mine and haven’t yet filled it up.

As far as I know, the issue arises only when there is a memory capacity of 64GB or more.



Setting amdgpu.sg_display=0 seemed to work for me.

PopOS 22.04
2x 1440p monitors
3.03 BIOS

White flickering was really bad before and I needed to pretty much restart my computer to buy me another couple hours before it happened again.

Another datapoint from me. Left the machine off overnight again but this time Slack wasn’t open but Alacritty still was (which I thought might have been the culprit as the only obviously gpu accelerated app I had open the first time around). Opened fine this morning-- I think someone mentioned a theory about memory pressure / electron / node. May be onto something there.

1 Like

I’ve been able to associate the flickering with AMD-Vi. I have the suspicion that this is an iommu grouping issue. I’ve had similar experiences with the root cause on a previous Intel desktop motherboard. That issue was resolved with a bios update.

Having captured the logs a few times successfully I reported the flickering issue on the amdgpu drm bug-tracker.

With the suggested 6.7-rc2 and rtc_cmos.use_acpi_alarm=1 I do not get any acpi errors on boot. Package wattage sits around 2.7w when idle with the total system power consumption around 5-6watts and at 8watts when the wifi chip is transmitting data.

When I re-awake from suspend I do get one error, however the logs are much quieter on this kernel. I’m going to run this rc for a few days and see how it goes.


Do keep us posted, we are tracking this actively.


Enabling UMA_Game_Optimized worked for me.

Arch on Gnome (Wayland) and Hyprland
Docks/Hubs: Literally any, displaylink and not, USB3.2 and USB4, dp alt mode and not.
Bios: 03.03
Kernel: 6.6.3-arch1-1

Enabling UMA_Game_Optimized worked fine for normal productivity, but when I ran a game for a while, the issue reappeared in the middle of the gaming session. Disabled scatter-gather and no more issues, but curious how much performance (or efficiency) is lost from this.

Other datapoint: like others, this issue often pops up whenever I make an app fullscreen in GNOME.

However, and strangely enough, the issue does not affect fullscreen Emacs windows. One might think that this relates to Emacs running under X (or xwayland), but this is not the case, since I run the Wayland-native Emacs with pgtk support enabled. That is, I’ve installed Emacs from Arch Linux’s extra/emacs-wayland.

(6.6 kernel)