Unable to boot Tumbleweed with kernel-firmware-amdgpu newer than 20250916-1.1

English Install/Boot/Login
1

Relevant Hardware:

  • MB: Gigabyte B650 Eagle AX rev 1.1
  • CPU: AMD Ryzen 9700X
  • dGPU: AMD RX 480 (MSI)

Currently running openSUSE Tumbleweed 20260329, kernel 6.19.10 (default)

I’ve been having this problem since ~October 2025, which is when I tried to zypper dup and my system wouldn’t boot anymore.
I’ve narrowed it down to kernel-firmware-amdgpu and had that in my zypper locks since then, and had no further issues keeping up with Tumbleweed.
Now I’d like to run a newer version of amdgpu, but I am still facing the same issue.

When I upgrade the kernel-firmware-amdgpu package to anything newer than 20250916-1.1, my system does not boot (hangs on a black screen, not even SYSRQ works) - unless I fully disable the “Raphael” iGPU through the UEFI. Having it enabled, even when all displays are plugged into the dGPU and “Initial display output” is set to “PCIE”, the system just hangs on boot. (Unless I use nomodeset, but not using the GPU doesn’t solve the problem)

What I’ve tried:

  • dracut/Regenerating initramfs
  • Enabling/disabling amdgpu Overdrive
  • Upgrading kernel-firmware-amdgpu separately, or with the other packages
  • Making a bootable flash drive from an original openSUSE Tumbleweed Snapshot .iso (didn’t boot either)

All of those did not help. Again, going back to kernel-firmware-amdgpu from mid-September 2025 through snapper Snapshots makes everything work, iGPU included. Displays plugged into MB/iGPU, programs running on it rather than dGPU, no issues.

I’m struggling to even get a clear reason/error with the newer package. The journalctl logs don’t show anything related to amdgpu except the usual, e.g. “Overdrive is enabled, please disable it before reporting any bugs unrelated to overdrive”

Disabling plymouth, the system hangs on the line “amdgpu: ATOM BIOS : 102-RAPHAEL-008”, before the screens all go dark and nothing further happens. The same line appears when booting with the old amdgpu package, but there it’s the last thing I see before sddm.

Plugging everything into the dGPU and leaving the iGPU on, the system hung on

amdgpu 0000:10:00.0: [drm:amdgpu_ring_test_helper [amdgpu]] *ERROR* ring vcn_dec_0 test failed (-110)
amdgpu 0000:10:00.0: amdgpu: hw_init of IP block <vcn_v3_0> failed -110
amdgpu 0000:10:00.0: amdgpu: amdgpu_device_ip_init failed
amdgpu 0000:10:00.0: amdgpu: Fatal error during GPU init

About which I can’t seem to find much information.

Feel free to ask for more information, I can reproduce this boot failure at any time just by enabling my iGPU/upgrading kernel-firmware-amdgpu.

Anything else I can try on my end? Is this something for the openSUSE Bugzilla?

Or the amdgpu one? But if this isn’t an openSUSE issue, are there really so few systems running Zen 5 with the iGPU enabled?

2

i probably wouldn’t be able to help you but this probably a regression
you could try to enter more deeper debugging mode so you can see more errors
there is kernel flags to use to do that i don’t know what will be helpful to you i
think i would start by having the highest verbosity i can

https://docs.kernel.org/admin-guide/kernel-parameters.html
https://wiki.archlinux.org/title/AMDGPU#Troubleshooting

this more for devolpers
https://docs.kernel.org/gpu/amdgpu/debugfs.html

1 Like
3

Thank you for giving me some pointers - interesting read.
Though as usual I found the solution to the months-old issue one day after asking: Unstable iGPU undervolt :man_facepalming:

Listed as “GFX Curve Optimizer” in the UEFI, I had it set to Negative 30. (Which, again, works flawlessly on 20250916-1.1 with the iGPU doing compositing, KDE, Firefox, decoding videos, etc.)

Took that out, enabled iGPU, system boots.
Fiddled with more UEFI settings, set everything back to the way it was yesterday (except GFX curve optimizer), system works.

GFX curve optimizer back to Negative 30, same vcn_dec_0 error as above.
Negative 24 gets to a text-based login where amdgpu is spamming complaints ( I'm not done with your previous command) and locks up soon after.

Negative 16 works, iGPU is enabled, visible and usable and kernel-firmware-amdgpu is still up-to-date.

Not sure if/how that has survived multiple UEFI updates but it’s entirely possible I’ve been restoring that along with all the other settings on Autopilot.
Either way, my mistake, case closed.

1 Like
4

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.