AI 300 Display locking up incessantly with Fedora 42 Xfce

Subject pretty much says it all. I log in and after < 10 minutes the display will lock up. I can move the mouse but that’s all. I can ssh into the laptop and reboot it but otherwise the display is unresponsive.

I can’t see anything in dmesg, systemd journal, or Xorg.log.

Has anyone experienced anything like this with this or prior AMD laptops and know of a workaround or where I might find more info on what is causing the crash?

Which Linux distro are you using?
Fedora

Which release version?
42, fully updated

Which kernel are you using?
kernel-6.14.2-300.fc42.x86_64

Which BIOS version are you using?

Which Framework Laptop 13 model are you using? AMD Ryzen™ AI 300

Try a usb keyboard and usb mouse.
That will narrow down the problem a bit.
That will determine if it is an input cover problem or something else.

I’m cautiously optimistic that it is just the Xfce compositor. I tried disabling it and was able to go for awhile without freezing.

A google search suggested the default vblank mode can cause problems with AMD GPUs and that xpresent supposedly works pretty well. So reenabled the compositor, ran xfconf-query -c xfwm4 -p /general/vblank_mode -t string -s "xpresent" --create, rebooted and so far it hasn’t frozen.

I’ll give it a few hours before I celebrate, but given before it was 5-10 minutes tops I’m hopeful.

I think this is all it was. I’ve been setting up for a couple hours and no more issues.

Does this happen with a Wayland compositor? It sounds like a bug with X11 to me.

Wait, AI 300 has shipped? DIY or pre built?

I’m guessing it’s not a problem with Wayland. I’ve only used Xfce for quite awhile which still relies X11.

Yes. DIY, Batch 1, it arrived yesterday.

Huh alright. I’m batch 1 for the board but haven’t been billed yet.

Glad you figured out the issue, enjoy!

1 Like

Could you install and check Wayland? I would like to confirm that it’s really an X11 bug. Additionally can you experiment with disabling PSR/PR/PSR-SU? Try these 3 options separately and share your results.
amdgpu.dcdebugmask=0x610
amdgpu.dcdebugmask=0x400
amdgpu.dcdebugmask=0x10

I set the vblank mode back to auto and confirmed the problem was back. After that I tried each of these in order. 0x610 appeared to work, 0x400 did not - I got a freeze pretty quickly, and I’ve been running now for just about 100 minutes with 0x10 so it also appears to work.

Give me a bit and I’ll install gnome desktop and try wayland without any values set to see what happens.

I am currently experiencing the exact same issue in Solus Budgie (X11) with an AI 300. I will try some of the above and report back.

Removed amdgpu.dcdebugmask=0x10 and restarted, issue came back..

Installed gnome and hit gnome-shell 48-alpha SEGV in cogl_texture_get_context (#3933) · Issues · GNOME / mutter · GitLab

Rather than work my way through that I uninstalled gnome, installed KDE Plasma and no issue, so as far as I can tell it is only a problem with X11.

Can you guys please open an issue in drm/amd please? What’s going on is some sort of incompatibility with PSR and X11. Setting vblank is working around it (PSR never engages).

We are potentially going to modify the driver to detect X11 and disable PSR, but haven’t decided. More data points on the prevalence of this problem will be helpful for that decision.

Btw - you will get better battery life in Wayland as a result of this.

How did the Phoronix’ review with this model on Ubuntu 25.04 not have any mention of this particular issue?

Is it because of the Mesa driver?

Also, how is it that this bug only needs to be filed ‘now’…instead of during internal testing by Framework?

That review isn’t with X11. X11 gets very little testing.

1 Like

I see. So even with Framework joining late to the AI 300 party, no one in the non-Framework wild even filed the bug?

I started a new issue:

3 Likes

Thanks for tracking this down! I have the same issue on FW13 Ryzen HX370, using Ubuntu 25.04 with X11 (I disabled Wayland for reasons, with WaylandEnable=false in /etc/gdm3/custom.conf) and was pulling my hair a bit. Using kernel 6.14.0-15-generic; if any of that helps.

I updated /etc/default/grub with:

GRUB_CMDLINE_LINUX_DEFAULT=“quiet splash amdgpu.dcdebugmask=0x610”