[SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

bernd32 · 2026-01-13 09:47:03

My current specs:
OS: Arch Linux
PROCESSOR: Intel Core i9-9900K
GRAPHICS: Gigabyte AMD Radeon RX 9070 XT 16GB GAMING OC
MONITORS: (1) 1080p@100hz; (2) 1440p@166hz (primary); (3) 1080p@60hz
MOTHERBOARD: ASRock Z370 Pro4 (supports up to pcie 3.0), BIOS Version: P4.30E
MEMORY: 16GB, DDR4
PSU: 850 W
Desktop: KDE Plasma 6.5.4, wayland session
Driver and kernel versions:
mesa 1:25.3.3-2
linux-lts 6.12.65-1
linux 6.18.5.arch1-1

A few days ago, I bought a new GPU: Gigabyte AMD Radeon RX 9070 XT 16GB GAMING OC. When I try to launch most games on Steam, they freeze after about ~5 seconds, showing either a black or a white screen with the following logs in journalctl:

Jan 13 13:37:02 archpc steam[2717]: 01/13 13:37:02 Init: Installing breakpad exception handler for appid(gameoverlayui)/version(1.0)/tid(4476)
Jan 13 13:37:05 archpc steam[2717]: Adding process 4483 for gameID 1245620
Jan 13 13:37:06 archpc steam[2717]: Setting breakpad minidump AppID = 1245620
Jan 13 13:37:06 archpc steam[2717]: SteamInternal_SetMinidumpSteamID: Caching Steam ID: 76561198024581306 [API loaded no]
Jan 13 13:37:06 archpc steam[2717]: WARNING: radv is not a conformant Vulkan implementation, testing use only.
Jan 13 13:37:06 archpc steam[2717]: Fossilize INFO: Overriding serialization path: "/mnt/SteamLib/SteamLibrary/steamapps/shadercache/1245620/fozpipelinesv6/steamapprun_pipeline_cache".
Jan 13 13:37:07 archpc plasma_waitforname[2629]: WaitForName: Service was not registered within timeout
Jan 13 13:37:07 archpc systemd[996]: dbus-:1.1-org.kde.KSplash@1.service: Main process exited, code=exited, status=1/FAILURE
Jan 13 13:37:07 archpc systemd[996]: dbus-:1.1-org.kde.KSplash@1.service: Failed with result 'exit-code'.
Jan 13 13:37:07 archpc steam[2717]: 01/13 13:37:07 minidumps folder is set to /tmp/dumps
Jan 13 13:37:07 archpc steam[2717]: 01/13 13:37:07 Init: Installing breakpad exception handler for appid(gameoverlayui)/version(20251223004807)/tid(4557)
Jan 13 13:37:07 archpc steam[2717]: 01/13 13:37:07 Init: Installing breakpad exception handler for appid(gameoverlayui)/version(1.0)/tid(4557)
Jan 13 13:37:17 archpc steam[2717]: reaping pid: 4476 -- gameoverlayui
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: Dumping IP State
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: Dumping IP State Completed
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: [drm] AMDGPU device coredump file has been created
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: [drm] Check your /sys/class/drm/card1/device/devcoredump/data
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: ring comp_1.1.0 timeout, signaled seq=1, emitted seq=3
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: Process eldenring.exe pid 4483 thread vkd3d_queue pid 4561
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: Starting comp_1.1.0 ring reset
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: reset compute queue (1:1:0)
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:32 vmid:0 pasid:0)
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: in page starting at address 0x0000000000000000 from client 10
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: GCVM_L2_PROTECTION_FAULT_STATUS:0x00040A40
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: Faulty UTCL2 client ID: CPC (0x5)
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: MORE_FAULTS: 0x0
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: WALKER_ERROR: 0x0
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: PERMISSION_FAULTS: 0x4
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: MAPPING_ERROR: 0x0
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: RW: 0x1
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: amdgpu: Ring comp_1.1.0 reset succeeded
Jan 13 13:37:18 archpc kernel: amdgpu 0000:03:00.0: [drm] device wedged, but recovered through reset
Jan 13 13:37:18 archpc steam[2717]: WARNING: radv is not a conformant Vulkan implementation, testing use only.
Jan 13 13:37:18 archpc steam[2717]: radv/amdgpu: The CS has been cancelled because the context is lost. This context is guilty of a hard recovery.
Jan 13 13:37:35 archpc steam[2717]: pid 4355 != 4354, skipping destruction (fork without exec?)
Jan 13 13:37:35 archpc kwin_wayland[1159]: QProcess: Destroyed while process ("/usr/lib/kwin_killer_helper") is still running.
Jan 13 13:37:35 archpc steam[2717]: Game Recording - game stopped [gameid=1245620]
Jan 13 13:37:35 archpc steam[2717]: Removing process 4483 for gameID 1245620

I've tried to launch several different games on Steam, and the only one in my Steam library that runs without issues is Battle for Wesnoth, which is a Linux native game. The unigine-superposition benchmark from the Phoronix Test Suite also runs without any problems on the linux 6.18.5.

I've tried various kernel parameters, different game launch options, and multiple Proton versions, but none of that helped. The only thing that did fix the issue was installing and booting into the linux-lts 6.12.* kernel, despite that all the AI chatbots advised me to update the driver and kernel to newer versions (yes, all AI advices were useless). I've played games for around 10 hours on the LTS kernel without any crashes or freezes, or any problems at all, which makes me think the GPU itself is fine.

I also found an Arch Wiki section that describes the exact same error messages I'm seeing in the logs:
https://wiki.archlinux.org/title/AMDGPU … ple_issues

However, I'm hesitant to mess with GPU overclocking or undervolting, especially since that section is marked with the note: "The factual accuracy of this article or section is disputed."

I don't have high expectations, but I'm hoping someone else has encountered a similar issue and can offer some advice. I'm also not sure whether sticking with the LTS kernel is a proper long-term solution or just a workaround. Does anyone have any ideas as to why these issues could be occurring with the current kernel version? Considering that a large number of Linux users use this GPU without problems on an actual kernel version, I can only assume that the software error is caused by an unfortunate combination of my hardware.

edit: fixed the problem with updating kernel to 6.18.6-arch1-1

Last edited by bernd32 (2026-01-22 18:59:33)

Luciddream · 2026-01-13 10:13:53

It's probably related to this bug https://gitlab.freedesktop.org/drm/amd/-/issues/4765 - a fix should be available soon

Are you running anything that is using ROCm? Maybe ollama in the background for example.

bernd32 · 2026-01-13 10:58:21

Luciddream wrote:

It's probably related to this bug https://gitlab.freedesktop.org/drm/amd/-/issues/4765 - a fix should be available soon
Are you running anything that is using ROCm? Maybe ollama in the background for example.

Hmm, I don't think so. I've tried closing most of the background user programs, like Firefox or SVP4, but that didn't solve the problem. Also I don't have ollama on my system. So the only AI-related app on my system is KataGo, but I think it has nothing to do with the problem, and it is not running in the background.

Luciddream · 2026-01-13 11:08:26

Can you try setting

RADV_DEBUG=nocompute

to see if it makes any difference?

Last edited by Luciddream (2026-01-13 11:08:54)

bernd32 · 2026-01-13 11:35:35

Luciddream wrote:

Can you try setting
RADV_DEBUG=nocompute
to see if it makes any difference?

Interesting: after rebooting into the current 6.18.5-arch1-1 kernel and setting this environment variable, the games (or at least 2 of them: Elden Ring and Mount&Blade2) now launch without any issues. Why is that? Is this a solution to the problem or is it just a workaround? Anyway, thank you!

Luciddream · 2026-01-13 11:38:54

bernd32 wrote:

Interesting: after rebooting into the current 6.18.5-arch1-1 kernel and setting this environment variable, the games (or at least 2 of them: Elden Ring and Mount&Blade2) now launch without any issues. Why is that? Is this a solution to the problem or is it just a workaround? Anyway, thank you!

Yes, just a workaround. It seems it triggers the same bug. Someone suggested this workaround on the issue tracker so it's cool that it works. I guess it won't be necessary in a couple of days when the proper fix lands in mesa

Lone_Wolf · 2026-01-13 12:28:36

The commit that caused this is in the kernel and the workaround patch is also for the kernel.

linux 6.19 development is in rc5, if this isn't backported to stable versions it's possible it will only land in the 6.20/7.0 kernel .
(You may have to build your own kernel )

Luciddream · 2026-01-13 12:31:09

Oops, you are correct, this is what I get for commenting while at work, mind stops thinking

bernd32 · 2026-01-13 12:33:33

Luciddream wrote:

I guess it won't be necessary in a couple of days when the proper fix lands in mesa

Let's hope for the best and wait for the developers to resolve the issue as soon as possible By the way, maybe you know if there are major trade-offs or side effects when using this 'RADV_DEBUG=nocompute' variable?

Lone_Wolf · 2026-01-13 12:51:55

The compute queue it disables can only be used through the vulkan compute api .

Disabling it is often suggested when troubleshooting vulkan graphical issues.

Comparing vulkaninfo output with & without this flag may shed some light on what's missing with the compute que disabled.

Luciddream · 2026-01-20 19:21:51

bernd32 wrote:

Let's hope for the best and wait for the developers to resolve the issue as soon as possible By the way, maybe you know if there are major trade-offs or side effects when using this 'RADV_DEBUG=nocompute' variable?

This is now fixed in the kernel in core-testing repository. You can either update or wait for it to land on the core repository.

Last edited by Luciddream (2026-01-20 19:26:07)

bernd32 · 2026-01-21 07:28:32

Glad to hear good news. I think I'll just wait since LTS kernel works fine at the moment.

bernd32 · 2026-01-22 18:20:54

UPDATE.

My system NVMe drive suddenly died today, so I had to buy a new one and reinstall Arch Linux. Before restoring my backup files, I decided to install Steam and run the games I previously had trouble launching on the 6.18.x kernel, and... they started without any issues. There were also no errors or even warnings in journalctl. I'm glad but also kind of confused.

I guess I need to be careful with restoring configuration files from backup now, because there is a chance that my problem was related to the system configuration? Or maybe it is because of the nvme drive? I don't know.

archpc% uname -r
6.18.6-arch1-1

I'm not sure if I should mark my thread as [solved]. I've solved the problem, but the root cause hasn't been identified yet.

Last edited by bernd32 (2026-01-22 18:24:46)

Luciddream · 2026-01-22 18:22:06

the kernel got in the core repository yesterday, so you probably updated yesterday or today. Sorry to hear about your NVME drive

Last edited by Luciddream (2026-01-22 18:22:28)

bernd32 · 2026-01-22 18:30:34

Luciddream wrote:

the kernel got in the core repository yesterday, so you probably updated yesterday or today. Sorry to hear about your NVME drive

I thought the fix patch would only be available starting with 6.19.

Luciddream wrote:

Sorry to hear about your NVME drive

No big deal I have daily backups of all my files on another computer.

Luciddream · 2026-01-22 18:39:31

bernd32 wrote:

I thought the fix patch would only be available starting with 6.19.

It should, but it has been backported. Here you can see the patches that are applied.

bernd32 · 2026-01-22 18:43:28

Luciddream wrote:

bernd32 wrote:
I thought the fix patch would only be available starting with 6.19.
It should, but it has been backported. Here you can see the patches that are applied.

Ok, great. In that case, I can mark the thread as solved.

Arch Linux

#1 2026-01-13 09:47:03

[SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#2 2026-01-13 10:13:53

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#3 2026-01-13 10:58:21

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#4 2026-01-13 11:08:26

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#5 2026-01-13 11:35:35

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#6 2026-01-13 11:38:54

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#7 2026-01-13 12:28:36

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#8 2026-01-13 12:31:09

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#9 2026-01-13 12:33:33

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#10 2026-01-13 12:51:55

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#11 2026-01-20 19:21:51

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#12 2026-01-21 07:28:32

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#13 2026-01-22 18:20:54

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#14 2026-01-22 18:22:06

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#15 2026-01-22 18:30:34

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#16 2026-01-22 18:39:31

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

#17 2026-01-22 18:43:28

Re: [SOLVED] AMDGPU crashing right after launching games on Linux 6.18.*

Board footer