Getting Occasional Graphical Bugs / Artifacting After Switching to AMD

I decided to main Linux after having a dual boot for a while, and I recently switched to an AMD card after having only Nvidia GPUs for a while just to make things simpler for myself. That switch was about 3 weeks ago. It was fine until about 3 or 4 days ago when I started getting some weird horizontal static-y lines / artifacting.

I’m hoping it’s not a sign that my new GPU is defective and needs to be replaced already. I noticed it doesn’t seem to happen on Windows, though on I haven’t had a chance to use Windows as much since this issue started and on Linux it only seems to happen sporadically. Sometimes I’ll go 10-20 minutes or more without a single artifact which only lasts a split second anyways.

The first time I noticed it was when I was playing a game using Wine, so I thought it was the game causing it, then I noticed it happen when I was just using Firefox. Like I said I’ve only had this new GPU for about 3 weeks, but this issue only started a few days ago so I’m not sure why. I recently switched to the Fedora 41 beta (fresh install) and after a few days I noticed the issue. Switched back to Fedora 40 (another fresh install) to see since it was fine before, but then it happened again. Any advice would be appreciated.

My setup:

OS: Fedora 41 (Gnome & Wayland) / Fedora 40 (Gnome & Wayland), Windows 11

Display 1: 1440p 144Hz Displayport 1.4 monitor to my right with Gsync only.

Display 2: 4K 120Hz HDMI 2.1 LG OLED C3 TV that I sometimes game on, has Freesync and Gsync support. For this I have to use a 50 foot fiber cable, not sure if the cable could have gotten damaged, hopefully not as it would be expensive to replace.

CPU: AMD 7800X3D
GPU: AMD Radeon RX 7900 XTX (Sapphire Nitro+ )
RAM: G.SKILL 32GB 6000MT/s
PSU: 1200W Corsair

I tried using screen capture to get these, but they don’t show up using the built in Gnome screen recorder. I had to just point my phone at the screen and record as I used the PC. This was a game running on Wine on Fedora 41. I have some video capture as well, not sure the best way to upload that. Like I said these artifacts last a split second and only happen maybe once ever 10-20 minutes or more.

Edit: It could also be related to the issue AMD has with HDMI 2.1 since I’m running on a 4K resolution with 120Hz and possibly 10 bit color, I’m not sure how that last one works on Fedora or Linux if it’s done automatically or if I have to pick that option somewhere like on Windows. I assume that is past the bandwidth that HDMI can handle with AMD on Linux. But I also once noticed the artifacting happening on my 1440p 144Hz monitor as well while both were connected and I was mainly using the 4K TV. I also once disconnected the 1440p monitor to see if it was a dual screen issue, but it still happened on the 4K TV.



I don’t know if it might be related, but someone else recently posted that kernel versions greater than 6.10.9 were exhibiting amdgpu driver problems on their system.

Can you boot to a kernel version less than or equal to 6.10.9 and test if that resolves the problem you are seeing?

2 Likes

Since I just did a fresh install of f40, I have kernel 6.8.5-301.fc40.86_64 . I can try that for a while.

Do you know when 6.10.9 was available for download on f40 / f41 ? I want to see if it kind of likes up with when I started noticing this or if it was before. Because like I said I’ve had this gpu for about 3 weeks and the issue only started like 4-5 days ago now.

Is there any way to tell by looking at the kind of artifacting if It’s likely to be software related, or if it’s a defective GPU that needs to be replaced? Since I just got it I’d like to know as soon as possibly because it would probably be easier exchanging it at the store than trying to do it through warranty.

It looks like 6.10.9 was built for Fedora Linux 40 on 08 Sep 2024: kernel-6.10.9-200.fc40 | Build Info | koji There may have been a few days delay before that update would have been visible to you due to the time it took the build to get through the testing/gating process and depending on how frequently your closest Fedora Linux mirror resynchronizes with the upstream repos. Also, your local package installer has a cache of the package metadata that might introduce a small delay between when the package shows up in the mirror and when you see it as being available for installation.

Edit: I think you should be able to find out exactly when you installed a specific kernel with dnf history list kernel followed by dnf history info <id>. (Or rpm -q --last kernel might be simpler if you still have that kernel installed.)

Someone who is more familiar with those types of errors might be able to tell the difference. I cannot.

The artifacting in the screenshots looks like HDCP snow and display connectivity instability.

I’ve seen HDMI instability on Polaris at 4K@60Hz, and suspect AMD still has issues at high-bandwidth stuff. An idea is to try creating a reduce-blank resolution to lower the bandwidth used.

I’m not sure what to try to make for 2 displays and DP, but here’s what I use:

video='HDMI-A-1:1920x1080MR@75'
  • HDMI-A-1 came from something and might be shown with proptest from Wayland (not xrandr). I think DP would be DP-1. Different ports on the GPU report something different so I’d get those for-sure from proptest.
  • M force-creates the specified res/refresh (overrides EDID)
  • R generates it reduce-blank (lower bandwidth req)

If you use GNOME on Xorg session you can also create the reduce-blank modelines and add/set em for quick testing with xrandr. This is what I did a while back when I had the RX 580s and 4K@60Hz:

cvt --reduced '3840' '2160' '60'
xrandr --newmode "3840x2160R"  533.00  3840 3888 3920 4000  2160 2163 2168 2222 +hsync -vsync
xrandr --addmode 'HDMI-A-0' '3840x2160R'
xrandr --output 'HDMI-A-0' --mode '3840x2160R'
1 Like

FWIW, the Arch wiki[1] suggests the following command to get a list of possible connector names:

for p in /sys/class/drm/*/status; do con=${p%/status}; echo -n "${con#*/card?-}: "; cat $p; done

  1. https://wiki.archlinux.org/title/Kernel_mode_setting#Forcing_modes ↩︎

1 Like

So I just got around to trying 6.8.5-301.fc40.86_64 and it happened again. Could this be something that got updated that isn’t part of the kernel? I’m using the Mesa Freeworld drivers to get hardware acceleration. Are there other things outside the kernel that recently got updated that could be causing this? Like I said when I try switching back to Windows 11 for a while I don’t notice the same artifacting, but I also didn’t have it for a bit over 2 weeks on Linux.

Edit: I also saw this point in the related posts at the bottom, and it looks similar to what I’m seeing.

I’m going to switch over to x11 for a while and see what happens, maybe it’s a Wayland issue.