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

  1. Usually started while on power, but would continue after unplugging
  2. Affected both internal display and external display when connected via USB-C
  3. Sorry, didn’t try with live-USB
  4. Definitely happened with just a single internal display

For me, I could reproduce by changing my screen resolution or zoom level. FYI I’m running the Fedora 39 beta KDE spin but it also happened with the regular Gnome Fedora 39 beta.

Adding the amdgpu.sg_display=0 kernel param has 100% fixed this for me - I haven’t seen this happen since adding that param.

1 Like

AMD has confirmed that this is difficult to duplicate, however, that using amdgpu.sg_display=0 is the way forward. Will be updating docs today for those affected.

Also if you can provide logs and steps to duplicate, please create an issue here:

2 Likes

Hi All,

I’ve just got my nice new shiny FW13AMD, and I’m super impressed, great work FW Team.

I’m a Debian fan, and yeh I know, not “officially” supported etc and this is a Fedora thread, but…

I’ve just loaded a fresh Debian 12(Bookworm) build, ver 12.2 and upgraded to kernel(from stable-backports) 6.4.0-0.deb12.2-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.4.4-3~bpo12+1 (2023-08-08) x86_64 GNU/Linux and can successfully replicate this “white screen” issue.

What I’ve been able to find is that when you suspend, then awake, if you have any application, like a browser etc, trying to access and use the 3D acceleration of the amdgpu driver, the screen corrupts the display output.

I simply close that browser(in my case Brave), and the corruption stops(in my case white solid screen or “techno” flickty flashes, glad I don’t have epilepsy conditions).

I hope that this might help/point us all in the right direction.

Let me know if there’s any other info you’d like.

Paul.

1 Like

Maybe a red herring, maybe not. I’ll throw this out here in case it helps. I have a 2015 15” MBP with a discrete AMD r9 GPU that I tried loading Debian 12 onto and got this exact same issue of white covering most of my screen, with blocks especially around text, making it unusable. The big difference is it happened 100% of the time. Only thing that machine has in common with FW 13 AMD is the AMD gpu I think, albeit a very different gpu. But perhaps this still points to the issue being with AMD’s gpu driver?

Instead of posting here it may be helpful to report this directly to AMD with logs. It seems like there is already an issue open: ubuntu 22.04, AMD 7840, SCREEN turn white/black/stripes (#2954) · Issues · drm / amd · GitLab. They probably need more information to be able to troubleshoot.

If it’s a different issue may be worth opening a new ticket so they know what is going on. No one from AMD is monitoring these forums from what I can tell.

There is an amd-gpu-firmware update in F39 updates-testing. Just upgraded and rebooted and no errors with full screen video after 5 min.

upgrading:
 amd-gpu-firmware                           noarch               20231030-1.fc39                   updates-testing
2 Likes

Okay, thanks @Jim_Williams - appreciate you calling this out. Please keep me posted! :slight_smile:

Over an hour now and no amd-gpu errors while running an external monitor off the hdmi expansion card. Looks promising!

I’ve installed these firmware updates, just rebooted, and am running 6.5.9-300 (both from fedora-updates-testing).

For others who would like to try:

[updates-testing]
name=Fedora $releasever - $basearch - Test Updates
#baseurl=http://download.example/pub/fedora/linux/updates/testing/$releasever/Everything/$basearch/
metalink=https://mirrors.fedoraproject.org/metalink?repo=updates-testing-f$releasever&arch=$basearch
enabled=1
includepkgs=kernel* kernel-modules* linux-firmware* *-firmware ### Add this line
countme=1
repo_gpgcheck=0
type=rpm
gpgcheck=1
metadata_expire=6h
gpgkey=file:///etc/pki/rpm-gpg/RPM-GPG-KEY-fedora-$releasever-$basearch
skip_if_unavailable=False

[...other sections here...]

The package in testing is amd-gpu-firmware-20231030-1.fc39 - Fedora Packages

I pulled apart the change notes and got:

  • Update to upstream 20231030 release
  • Update firmware file for Intel Bluetooth AX203/AX210/AX211/
  • Update firmware file for Intel Bluetooth Magnetor AX101/AX201/AX211
  • rtl_nic: update firmware of RTL8156B
  • Update AMD cpu microcode
  • amdgpu: update SMU 13.0.0 firmware
  • add Amlogic bluetooth firmware
  • i915: Add GuC v70.13.1 for DG2, TGL, ADL-P and MTL
  • iwlwifi: add a missing FW from core80-39 release
  • WHENCE: add symlink for BananaPi M64
  • i915: Update MTL DMC to v2.17
  • amdgpu: update various firmware from 5.7 branch
  • iwlwifi: add FWs for new GL and MA device types with multiple RF modules
  • amd_pmf: Add initial PMF TA for Smart PC Solution Builder
  • Update FW files for MRVL PCIE 8997 chipsets
  • rtl_bt: Update RTL8851B BT USB firmware to 0x048A_D230
  • iwlwifi: add new FWs from core81
  • iwlwifi: update cc/Qu/QuZ firmwares for core81-65 release
2 Likes

I’d like to report that I’m seeing the same type of artifacts (big white squares) on Debian Trixie (current testing, future Debian 13) with bios 3.03 and kernel 6.5.8 using Plasma 5.27.8.

So it seems the issue is not Fedora-specific.

While this is admittedly a cool effect, I was able to trigger this issue again with a full-screen application (Skyrim Special Edition) using non-windowed mode in native Steam.

This was with 6.5.9-300 and the 20231030 amd-gpu-firmware package.

This is with 32GB of Framework-supplied DDR5 RAM and no external monitors attached (although replicated with an external monitor as well). 2x USB-C, 2X USB-A expansion cards.

I am definitely seeing an improvement with KDE, 3.03, Linux 6.5.9, and the new 20231030 firmware. Unplugging and plugging back in a dock no longer causes a rapidly flashing white artifact. I am not using amdgpu.sg_display=0 either.

Very exciting to see this much progress.

Per AMD themselves, please try this and report back:

1 Like

If you are still experiencing this. As a test can you try to enable the bios item and report back if this helps or not:
Advanced->iGPU Configuration->UMA_GAME_OPTIMIZED

This will allocate more memory to the GPU.
I discussed this with AMD, and they suspect that these visual artifacts are caused by high memory allocation on the GPU. So providing more memory may help alleviate this issue.

5 Likes

This is actually a great idea and aligns well with what is described in the Fedora bugzilla thread.
In there they speculate that the issue is associated with near full vram usage.

If I can reliably reproduce that with the webgl fish demo I’ll report back.

Oh, maybe that is why I never experienced this. Certainly worth a try for the people affected! The only time I saw it was in a memory heavy game quite a while ago and I toggled that quite a while ago as well.

  • Happens regardless of power or battery
  • Internal display (have not tried an external display yet)
  • have not tested with live usb. (if this would be helpful don’t mind testing this.)

I’ve observed this on sway and i3. Most frequently occurs after waking from sleep if I did not shut the laptop lid myself. Once it occurs in a given boot, I can make it go away by killing sway or i3 (which ever I’m using)and restarting sway or i3. However The longer the laptop has been on the more frequently this occurs. around a week or so it becomes so frequent that I just reboot. Rebooting keeps it at bay for a while. (some times a day or two some times only a few hours) origionaly thought this was a wayland issue untill I tried switching to i3 to get rid of it (which failed, however the only time I got it to happen on x11 was by messing with the refresh rates or having it happen in sway and switching to i3 before killing sway). This happens at all the scales I’ve tried fractional or not. Though it seem to happen less frequently at 1.

It was alway in the back of my head to check the framework forums but I feel silly for not checking sooner.

I skimed this thread and I saw:

  • update the bios to 3.03 (but someone reported that made it worse)
  • Adding the amdgpu.sg_display=0 kernel param (is that an x11 specific?)
    Which is the correct fix?

Also do you still need people to file bug reports? And who do we file with, redhat, fedora, or amd? all have been mentioned in this thread.

All,

Please try what is suggested here and report back.

We’re actively tracking this and need your A/B testing to confirm if this is helping.

Trying to understand this…the A/B testing is to see if this “may help alleviate this issue.”…and not necessary a final fix? Correct?