Testing request: graphical corruption on Intel laptops on kernel 6.19

Hey folks!

There are a couple of proposed blockers for Fedora 44 Beta which seem to be hardware-specific, both on Intel graphics:

In deciding whether these should be release blockers, it would be helpful to know if they affect other hardware. If folks (who are comfortable with testing pre-release stuff) who have laptops with Intel graphics could test booting a Fedora 44 nightly or just a kernel 6.19 build from Koji and see if they’re affected by any graphical corruption / flickering / tearing, that’d be great. If you do have a problem, please test whether booting with i915.enable_psr=0 avoids it, and maybe whether it differs based on whether an external display is plugged in.

Thanks a lot!

Is it any particular “generation” of Intel graphics, or are all of them potentially affected?

XPS Plus 9320 was 12th/13th gen Intel, apparently. The systems from the other bug are Meteor Lake.

Cool. FWIW I tried Fedora 44 KDE nightly on my 8th gen iGPU-only laptop and I didn’t see any such graphical issues.

That’s probably of fairly low informational value though, if newer chips are the ones likely to be affected.

Affected user here!:sweat_smile:

Based on tests on my own devices and feedback from friends I know, I roughly speculate that it may be related to the following factors:

  1. Intel processors newer than MTL or ARL

  2. 120Hz high-refresh-rate screens, or screens that support Variable Refresh Rate (VRR)

I have also seen some other related bugs,

such as this one: github.com/CachyOS/linux-cachyos/issues/716

Some AMD users reported issues with VRR on external displays after kernel 6.19. While a subsequent fix worked for some users, it did not for others.

This bug likely affects newer machines more, so we may need users with new laptop to test it.We could also reach out to some OEMs for help.

I personally tried emailing Mark Pearson, Lenovo’s Linux technical support, to report this issue, but haven’t received a response yet.

Please provide your hardware specs, thanks.

ThinkPad P14s Gen5
Intel ARL-H 285H
14.5 inch (3072 x 1920) IPS panel with 120Hz

I’m affected as well.

I have a Lenovo Yoga Pro 9 16IAH10 with a Intel(R) Core™ Ultra 9 285H running on up-to-date Fedora 43 Cinnamon.

$ lspci -nn | grep -i vga
00:02.0 VGA compatible controller [0300]: Intel Corporation Arrow Lake-P [Arc Pro 130T/140T] [8086:7d51] (rev 03)

$ uname -r
6.19.13-200.fc43.x86_64

$ head /proc/cpuinfo  | grep name
model name	: Intel(R) Core(TM) Ultra 9 285H

My login screen would always freeze completely while typing my password. I had to force the system to use the xe driver (instead of the i915) and disable PSR. Using the boot options:

xe.enable_psr=0 xe.force_probe=7d51 i915.force_probe=!7d51

I was just going to say that modern Intel devices (like your laptop) are supposed to use more modern Intel XE driver which is specifically designed for them; i915 is obsolete technology.

I have the same platform (Intel 255H) and I’m on XE for about half a year already, everything works fine (currently using 6.19.13-200.fc43.x86_64).

by the way, as long as Fedora is known as “bleeding edge distro”, is there a reason to have i915 by default and not XE?

I also expected it to use the XE driver by default, but on a fresh Fedora install both i915 and XE are installed and initially both try to bind to the GPU. On my machine the i915 driver “wins” and takes control, which is why I had to add the boot options so that XE would actually bind to the device.

While I was trying to search the internet for help with this freezing issue I found this thread, and since @adamwill was explicitly asking for more affected Intel graphics users, I wanted to confirm that I’m also hitting a problem with PSR.

Given that the GPU is like already a year old (released january 2025), a complete freeze of the display right after a fresh install probably needs some attention.

Here’s how to activate XE on Fedora:

  1. check your GPU PCI ID:
    lspci -nn | grep VGA
    00:02.0 VGA compatible controller [0300]: Intel Corporation Device [8086:7d51]

  2. force XE driver:
    sudo grubby --update-kernel=ALL --args=“i915.force_probe=!7d51 xe.force_probe=7d51”

  3. reboot

  4. check if XE driver is indeed in use:

lsmod | grep -E “^(i915|xe)”

xe 4296704 109 ← Xe active
i915 5373952 0 ← i915 not used

Actually, this issue was already tracked down and fixed in a later kernel update. It’s related to PSR.

And honestly, I wouldn’t recommend turning on the Xe GPU driver unless you’re on supported hardware. I’m using an ARL-H U285H, and I tried enabling the Xe driver myself. It caused all sorts of problems—Ollama couldn’t read VRAM, games had weird rendering glitches, the whole mess. When I reported it to Intel, they basically said, “The Xe driver is only officially supported on LNL and newer integrated graphics. Yours might let you enable it, but it’s not covered.” Then they ignored me completely.

So unless you’ve got a B580 discrete GPU or an LNL or PTL iGPU, don’t expect Intel to care if you run into issues—even though you’ve technically been able to turn on the Xe driver since Tiger Lake.

Confirming this affects Arrow Lake-U (Core Ultra 7 265U, Intel Graphics 8086:7d41) on the Lenovo ThinkPad 21NX series with Fedora 44 kernel 6.19.14-300.fc44. Same Selective fetch area calculation failed in pipe A and ERROR Atomic update failure on pipe A signature. Filed at 2467676 – i915: "Atomic update failure on pipe A" / PSR2 selective fetch failure causes compositor freezes on Arrow Lake-U (Core Ultra 7 265U) with kernel 6.19.14-300.fc44.

Adding configuration detail in case it helps narrow this down on Arrow Lake-U:
The freezes occur with an external HDMI display connected (not the Dell XPS 9320 pattern, where attaching one suppresses the symptom). Both internal and external are running at 60 Hz, so the 120 Hz factor from the ThinkPad P14s case isn’t relevant here either.

  • Internal: eDP-1, 1920×1200 @ 60 Hz (Lenovo ThinkPad 21NX, Arrow Lake-U panel, eDP)
  • External: HDMI-A-1, 1920×1080 @ 60 Hz
  • Both pipes active at the time of the i915 errors.