AMDGPU crash with kernel freeze / instability on RX 5700 XT

Marichan · 2026-09-24 17:31:30

Hi everyone, having some issues on my RX 5700 XT, while (trying) to play certain specifics games (In this case Monster Hunter world)

After some time playing I started experiencing either full on kernel panics or complete crash of the graphical session, with only alternative ttys accessible and sometimes the graphic system remaining unstable (screen blacking out intermittently, multiple GPU resets triggering)

In general the kernel messages displayed:

kernel: amdgpu 0000:03:00.0: [gfxhub] page fault (src_id:0 ring:24 vmid:6 pasid:11)
kernel: amdgpu 0000:03:00.0:  Process kwin_wayland pid 1107 thread kwin_wayla:cs0 pid 1125
kernel: amdgpu 0000:03:00.0:   in page starting at address 0x0000800001c37000 from client 0x1b (UTCL2)
kernel: amdgpu 0000:03:00.0: GCVM_L2_PROTECTION_FAULT_STATUS:0x00600C30
kernel: amdgpu 0000:03:00.0:          Faulty UTCL2 client ID: CPG (0x6)
kernel: amdgpu 0000:03:00.0:          MORE_FAULTS: 0x0
kernel: amdgpu 0000:03:00.0:          WALKER_ERROR: 0x0
kernel: amdgpu 0000:03:00.0:          PERMISSION_FAULTS: 0x3
kernel: amdgpu 0000:03:00.0:          MAPPING_ERROR: 0x0
kernel: amdgpu 0000:03:00.0:          RW: 0x0
kernel: [drm:gfx_v10_0_bad_op_irq [amdgpu]] *ERROR* Illegal opcode in command stream
kernel: amdgpu 0000:03:00.0: Dumping IP State
kernel: amdgpu 0000:03:00.0: Dumping IP State Completed
kernel: amdgpu 0000:03:00.0: [drm] AMDGPU device coredump file has been created
kernel: amdgpu 0000:03:00.0: [drm] Check your /sys/class/drm/card1/device/devcoredump/data
kernel: amdgpu 0000:03:00.0: ring comp_1.3.1 timeout, signaled seq=1034, emitted seq=1036
kernel: amdgpu 0000:03:00.0:  Process MonsterHunterWo pid 4160 thread dxvk-submit pid 4209
kernel: amdgpu 0000:03:00.0: Starting comp_1.3.1 ring reset
kernel: amdgpu 0000:03:00.0: fail to wait on hqd deactivate
kernel: amdgpu 0000:03:00.0: Ring comp_1.3.1 reset failed
kernel: amdgpu 0000:03:00.0: GPU reset begin!. Source:  1
kernel: amdgpu 0000:03:00.0: BACO reset

*hard crash, log stops*

or

kernel: amdgpu 0000:03:00.0: [gfxhub] page fault (src_id:0 ring:24 vmid:7 pasid:19)
kernel: amdgpu 0000:03:00.0:  Process Xwayland pid 1261 thread Xwayland:cs0 pid 1272
kernel: amdgpu 0000:03:00.0:   in page starting at address 0x0000800003a8c000 from client 0x1b (UTCL2)
kernel: amdgpu 0000:03:00.0: GCVM_L2_PROTECTION_FAULT_STATUS:0x00700C30
kernel: amdgpu 0000:03:00.0:          Faulty UTCL2 client ID: CPG (0x6)
kernel: amdgpu 0000:03:00.0:          MORE_FAULTS: 0x0
kernel: amdgpu 0000:03:00.0:          WALKER_ERROR: 0x0
kernel: amdgpu 0000:03:00.0:          PERMISSION_FAULTS: 0x3
kernel: amdgpu 0000:03:00.0:          MAPPING_ERROR: 0x0
kernel: amdgpu 0000:03:00.0:          RW: 0x0
kernel: amdgpu 0000:03:00.0: Dumping IP State
kernel: amdgpu 0000:03:00.0: Dumping IP State Completed
kernel: amdgpu 0000:03:00.0: [drm] AMDGPU device coredump file has been created
kernel: amdgpu 0000:03:00.0: [drm] Check your /sys/class/drm/card1/device/devcoredump/data
kernel: amdgpu 0000:03:00.0: ring gfx_0.0.0 timeout, signaled seq=2163117, emitted seq=2163120
kernel: amdgpu 0000:03:00.0:  Process Xwayland pid 1261 thread Xwayland:cs0 pid 1272
kernel: amdgpu 0000:03:00.0: Starting gfx_0.0.0 ring reset
kernel: amdgpu 0000:03:00.0: Ring gfx_0.0.0 reset failed
kernel: amdgpu 0000:03:00.0: GPU reset begin!. Source:  1
kernel: amdgpu 0000:03:00.0: BACO reset
kernel: amdgpu 0000:03:00.0: GPU reset succeeded, trying to resume
kernel: amdgpu 0000:03:00.0: [drm] PCIE GART of 512M enabled (table at 0x0000008000300000).
kernel: amdgpu 0000:03:00.0: VRAM is lost due to GPU reset!

*every time the grpahical interface tried to restart a new series of reset messages would be triggered*

While trying to understand if this was comparable to similar issues already happening, I found out that:

this issue can be changed from sporadic to replicable by running the game in widowed mode. It is so bad that sometimes it will crash the second the 3D title screen is loaded.
the hard crash and the Illegal opcode in command stream error are not correlated
launching the game with the RADV_QUEUE_DISABLE=compute envarg fixes it (at least the easy to reproduce crash setup), albeit with the loss from 90fps to 60fps with very low lows

I am fully up to date, and in general the rest of the system shows no problems. I have tried lowering clocks but of course it does nothing. crash is unaffected by purging the shader cache. Using plasmashell on wayland, no plugins, i7-4790K CPU.

Seems to me an issue with queue synchronizations on context switches, tho I am not sure if I should report this to mesa, proton or directly the amd team.

I haven't been able to test other big games that might make use of compute pipelines due to having lost my steam library to an hard drive crash, tho I am in the process of redownloading some. Similarly I could potentially test if this happens on my old polaris GPU. However I could not get any other game to trigger this crash, including emualtion of a certain recent nintendo console which I am pretty sure the emulator uses compute shaders

Lone_Wolf · 2026-09-25 10:51:17

GCVM_L2_PROTECTION_FAULT_STATUS

Nasty error and hard to troubleshoot, often connected with compute on specific chipsets.

If you want help investigating this upload a coredump and full journal of a boot where this occurred.

Marichan · 2026-09-25 19:12:16

Yeah, seems a mess, at least disabling the compute pipeline helps but maybe it can help making the drivers better in the future.

Anyway I managed to recover one of the device coredump by re-triggering one:
https://paste.c-net.org/AthertonTravers
and this is the associated journalctl, it's stupid long since i was testing other games before, the crash happens around line 7979, and it was a partial recover one:
https://paste.c-net.org/TerranceCritic

Lone_Wolf · 2026-09-26 11:22:47

amdgpu.si_support=1 radeon.si_support=0 amdgpu.cik_support=1 radeon.cik_support=1

Those options should only be needed for gcn videocards, your RX 5700 XT is several generations newer then that.
Please remove them.

MSI MS-7923/Z97 GUARD-PRO (MS-7923), BIOS V1.10 02/16/2016

https://www.msi.com/Motherboard/Z97-GUARD-PRO/support doesn't list 1.10 but shows 1.A at almost the same date.
Since hexadecimal A equals decimal 10 it's highly likelky it's the same firmware version .

Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: CPU 0: Machine Check: 0 Bank 4: fe00000000800400
Sep 24 18:21:37 ArchHome kernel: fbcon: Taking over console
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: TSC 0 ADDR ffffffff9e2b34b9 MISC ffffffff9e2b34b9 
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: PROCESSOR 0:306c3 TIME 1790266890 SOCKET 0 APIC 0 microcode 0
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: CPU 1: Machine Check: 0 Bank 3: be00000000800400
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: TSC 0 ADDR ffffffff9f11967f MISC ffffffff9f11967f 
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: PROCESSOR 0:306c3 TIME 1790266890 SOCKET 0 APIC 2 microcode 28
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: CPU 2: Machine Check: 0 Bank 3: be00000000800400
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: TSC 0 ADDR ffffffff9e2b34b9 MISC ffffffff9e2b34b9 
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: PROCESSOR 0:306c3 TIME 1790266890 SOCKET 0 APIC 4 microcode 28
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: CPU 3: Machine Check: 0 Bank 3: be00000000800400
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: TSC 0 ADDR ffffffffc07bc07e MISC ffffffffc07bc07e 
Sep 24 18:21:37 ArchHome kernel: mce: [Hardware Error]: PROCESSOR 0:306c3 TIME 1790266890 SOCKET 0 APIC 6 microcode 28

There are no other MCE errrors in the log which suggests they're not related to the crashes , but they do need to be investigated.

Sep 24 23:42:21 ArchHome kernel: hid-steam 0003:28DE:1142.0002: steam_get_serial: invalid reply (ae 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00)
Sep 24 23:42:21 ArchHome kernel: hid-steam 0003:28DE:1142.0002: steam_get_attributes: invalid reply (ae 15 01 46 43 36 30 38 32 35 31 41 41 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00)
Sep 24 23:42:21 ArchHome kernel: hid-steam 0003:28DE:1142.0002: steam_register:steam_get_attributes failed with error -5
Sep 24 23:42:21 ArchHome kernel: hid-steam 0003:28DE:1142.0002: Steam Controller 'XXXXXXXXXX' connected
Sep 24 23:42:21 ArchHome systemsettings[6861]: INFO: Bad GET_ATTRIBUTES_VALUES response for controller 0x55bb11572480
Sep 24 23:42:21 ArchHome systemsettings[6861]: terminate called after throwing an instance of 'std::logic_error'
Sep 24 23:42:21 ArchHome systemsettings[6861]:   what():  basic_string: construction from null is not valid
Sep 24 23:42:21 ArchHome systemsettings[6861]: KCrash: Application 'systemsettings' crashing... crashRecursionCounter = 2
Sep 24 23:42:21 ArchHome systemd-coredump[51638]: Process 6861 (systemsettings) of user 1000 terminated abnormally with signal 6/ABRT, processing...
Sep 24 23:42:21 ArchHome systemd[1]: Created slice Slice /system/drkonqi-coredump-processor.
Sep 24 23:42:21 ArchHome systemd[1]: Created slice Slice /system/systemd-coredump.
Sep 24 23:42:21 ArchHome systemd[1]: Started Process Core Dump (PID 51638/UID 0).
Sep 24 23:42:21 ArchHome systemd[1]: Started Pass systemd-coredump journal entries to relevant user for potential DrKonqi handling.
Sep 24 23:42:23 ArchHome firefox[33588]: JavaScript warning: https://assets.aliexpress-media.com/g/a227ca/61F4nKXFqoL/0.0.4/f4e643dc9d425543fc645b3a01b9e0fa.js, line 1: WEBGL_debug_renderer_info is deprecated in Firefox and will be removed. Please use RENDERER.
Sep 24 23:42:25 ArchHome systemd-coredump[51639]: Process 6861 (systemsettings) of user 1000 dumped core.
                                                  
                                                  Stack trace of thread 6861:
                                                  #0  0x00007fb309a9a17c n/a (libc.so.6 + 0x9a17c)
                                                  #1  0x00007fb309a3e5d0 raise (libc.so.6 + 0x3e5d0)
                                                  #2  0x00007fb30c2ec659 _ZN6KCrash19defaultCrashHandlerEi (libKF6Crash.so.6 + 0x8659)
                                                  #3  0x00007fb309a3e6f0 n/a (libc.so.6 + 0x3e6f0)
                                                  #4  0x00007fb309a9a17c n/a (libc.so.6 + 0x9a17c)
                                                  #5  0x00007fb309a3e5d0 raise (libc.so.6 + 0x3e5d0)
                                                  #6  0x00007fb309a25685 abort (libc.so.6 + 0x25685)
                                                  #7  0x00007fb309e9ac54 n/a (libstdc++.so.6 + 0x9ac54)
                                                  #8  0x00007fb309eb581a n/a (libstdc++.so.6 + 0xb581a)
                                                  #9  0x00007fb309e9a5ed _ZSt9terminatev (libstdc++.so.6 + 0x9a5ed)
                                                  #10 0x00007fb309eb5ad6 __cxa_throw (libstdc++.so.6 + 0xb5ad6)
                                                  #11 0x00007fb309e9eeff _ZSt19__throw_logic_errorPKc (libstdc++.so.6 + 0x9eeff)
                                                  #12 0x00007fb300c14105 n/a (kcm_gamecontroller.so + 0xe105)
                                                  #13 0x00007fb300c1bfff n/a (kcm_gamecontroller.so + 0x15fff)
                                                  #14 0x00007fb30a401dab n/a (libQt6Core.so.6 + 0x201dab)
                                                  #15 0x00007fb30a413d9d _ZN6QTimer10timerEventEP11QTimerEvent (libQt6Core.so.6 + 0x213d9d)
                                                  #16 0x00007fb30a3f8e89 _ZN7QObject5eventEP6QEvent (libQt6Core.so.6 + 0x1f8e89)
                                                  #17 0x00007fb30b705a4f _ZN19QApplicationPrivate13notify_helperEP7QObjectP6QEvent (libQt6Widgets.so.6 + 0x105a4f)
                                                  #18 0x00007fb30a390538 _ZN16QCoreApplication15notifyInternal2EP7QObjectP6QEvent (libQt6Core.so.6 + 0x190538)
                                                  #19 0x00007fb30a587591 _ZN14QTimerInfoList14activateTimersEv (libQt6Core.so.6 + 0x387591)
                                                  #20 0x00007fb30a6c01b9 n/a (libQt6Core.so.6 + 0x4c01b9)
                                                  #21 0x00007fb3092e3cae n/a (libglib-2.0.so.0 + 0x6dcae)
                                                  #22 0x00007fb3092e5248 n/a (libglib-2.0.so.0 + 0x6f248)
                                                  #23 0x00007fb3092e5395 g_main_context_iteration (libglib-2.0.so.0 + 0x6f395)
                                                  #24 0x00007fb30a6bcdf2 _ZN20QEventDispatcherGlib13processEventsE6QFlagsIN10QEventLoop17ProcessEventsFlagEE (libQt6Core.so.6 + 0x4bcdf2)
                                                  #25 0x00007fb30a39f1a4 _ZN10QEventLoop4execE6QFlagsINS_17ProcessEventsFlagEE (libQt6Core.so.6 + 0x19f1a4)
                                                  #26 0x00007fb30a39400b _ZN16QCoreApplication4execEv (libQt6Core.so.6 + 0x19400b)
                                                  #27 0x000055bb02354811 n/a (systemsettings + 0x15811)
                                                  #28 0x00007fb309a27781 n/a (libc.so.6 + 0x27781)
                                                  #29 0x00007fb309a278b9 __libc_start_main (libc.so.6 + 0x278b9)
                                                  #30 0x000055bb02355ae5 n/a (systemsettings + 0x16ae5)

There have been cases where the failing compute queue was merely the first symptom instead of the cause of the GCVM_L2_PROTECTION_FAULT .
Can you try non-steam gamecontrollers / peripherals ?
wired standard devices like a 5 euro usb mice with just 3 buttons and a wheel give the best chance to determine if the controllers are involved.

Marichan · 2026-09-26 15:06:46

Those options should only be needed for gcn videocards, your RX 5700 XT is several generations newer then that.

yeah, I noticed as wel when uploading the log, I recently upgraded from a polaris card, I had forgot to rebuild grub config after removing them.

There are no other MCE errrors in the log which suggests they're not related to the crashes , but they do need to be investigated.

I noticed those errors as well, but they are present only in the boot after a crash. For example in the following log they are not present, since I was coming off a fresh boot.

Can you try non-steam gamecontrollers / peripherals ?

Here you go (https://paste.c-net.org/PfeifferSwirl, long log again, I have been using the pc to watch some videos), I pulled the dongle out (line 4168) one minute before starting with just kb / mice. Game crashes all the same. Also, if the computer was hard rebooted the steam controller would act unreliable until rebooted itself so probably the messages in the other log are about that.

Lone_Wolf · 2026-09-27 18:46:12

Can't prove it, but those 2 logs give me the impression that this may be wayland/xwayland specific .

Please test with a simple X environment that doesn't require gtk and/or qt.
a few options are : twm, openbox, i3 , icewm

czr · 2026-09-29 00:14:36

Created account to thank OP for reporting this and add some details regarding same exact crashes on RX 5700 XT.

I've been struggling with this for a year now since I updated PC and switched to Linux.

It took me weeks just to figure out that RADV_QUEUE_DISABLE=compute (previously RADV_DEBUG=nocompute) indeed fixes the issue - unfortunately with performance hit: 12fps avg less in Cyberpunk 2077 benchmark.

Setup / howto reproduce / things i've tried:

- latest kernel 7.*, os is up to date - the issue is not new, 1yr at least
- crashes happen in heavy dx12 games on Proton (e.g. both Horizon games, Marvels Spiderman 2, Indiana Jones)
- I use COSMIC desktop but this happens also on Plasma, I always use Wine-Wayland when playing

- crashes happen mostly during heavy workloads - in other words lots of stuff happening on screen
- after undervolting and downclocking the card looks like things improved - no crashes during action but immediately after when cutscene is loaded

I've also made sure that the card runs stable - no throttling of any kind. I've capped fps at 100. Things improved (sort off, still testing) but the issue is not resolved.

Hope this helps. Will be grateful for any tips / workarounds.

[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: [gfxhub] page fault (src_id:0 ring:24 vmid:5 pasid:115)
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0:  Process cosmic-comp pid 32251 thread cosmic-com:cs0 pid 32288
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0:   in page starting at address 0x00008000081fa000 from client 0x1b (UTCL2)
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: GCVM_L2_PROTECTION_FAULT_STATUS:0x00500C30
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 Faulty UTCL2 client ID: CPG (0x6)
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 MORE_FAULTS: 0x0
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 WALKER_ERROR: 0x0
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 PERMISSION_FAULTS: 0x3
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 MAPPING_ERROR: 0x0
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 RW: 0x0
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: [gfxhub] page fault (src_id:0 ring:221 vmid:0 pasid:0)
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0:   in page starting at address 0x00000000029d6000 from client 0x1b (UTCL2)
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: GCVM_L2_PROTECTION_FAULT_STATUS:0x000009BA
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 Faulty UTCL2 client ID: CPF (0x4)
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 MORE_FAULTS: 0x0
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 WALKER_ERROR: 0x5
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 PERMISSION_FAULTS: 0xb
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 MAPPING_ERROR: 0x1
[Tue Sep 29 01:29:01 2026] amdgpu 0000:03:00.0: 	 RW: 0x0
[Tue Sep 29 01:29:03 2026] amdgpu 0000:03:00.0: Dumping IP State
[Tue Sep 29 01:29:03 2026] amdgpu 0000:03:00.0: Dumping IP State Completed
[Tue Sep 29 01:29:03 2026] amdgpu 0000:03:00.0: [drm] AMDGPU device coredump file has been created
[Tue Sep 29 01:29:03 2026] amdgpu 0000:03:00.0: [drm] Check your /sys/class/drm/card1/device/devcoredump/data
[Tue Sep 29 01:29:03 2026] amdgpu 0000:03:00.0: ring gfx_0.0.0 timeout, signaled seq=21997885, emitted seq=21997888
[Tue Sep 29 01:29:03 2026] amdgpu 0000:03:00.0:  Process cosmic-comp pid 32251 thread cosmic-com:cs0 pid 32288
[Tue Sep 29 01:29:03 2026] amdgpu 0000:03:00.0: Starting gfx_0.0.0 ring reset
[Tue Sep 29 01:29:03 2026] amdgpu 0000:03:00.0: Ring gfx_0.0.0 reset failed

czr · 2026-09-29 21:27:54

**@Marichan** I'm still testing but looks like I've found a way to avoid this crash - approx 5h of uninterrupted gameplay with this settings:

- LACT - UV and underclock the card - In my case the sweet spot is GPU Pstate2 - 2000mhz @ 1107mV (this solved crashes during processing demanding gameplay)
- set amdgpu.ppfeaturemask=0xffff7fff in your bootloader (this avoids crashes during transitions when card downlocks and upclocks rapidly)
- capped fps at 100 (might be irrelevant)

My card is Sapphire Nitro+ RX 5700XT Black Edition. Hope this helps.

Last edited by czr (2026-09-29 21:29:21)

Marichan · 2026-09-30 00:37:49

czr wrote:

**@Marichan** I'm still testing but looks like I've found a way to avoid this crash - approx 5h of uninterrupted gameplay with this settings:
- LACT - UV and underclock the card - In my case the sweet spot is GPU Pstate2 - 2000mhz @ 1107mV (this solved crashes during processing demanding gameplay)
- set amdgpu.ppfeaturemask=0xffff7fff in your bootloader (this avoids crashes during transitions when card downlocks and upclocks rapidly)
- capped fps at 100 (might be irrelevant)
My card is Sapphire Nitro+ RX 5700XT Black Edition. Hope this helps.

I tried full well to use LACT, this is actually how I found the reliable emthod to trigger the crash, by having the desktop showing. Undervolting, avoiding throttling and so on helps with the game and temps but I will still get the crash reliably. I also already tried the ppfeaturemask, no dice, and btw you might not haven't set it properly if you are using LACT and setting this in your bootloader. The ppfeaturemask is used to enable overdrive as well and lact will create a /etc/modprobe.d/99-amdgpu-overdrive.conf that overrides the boot option with all bits sets, so make you're not in this situation by catting /sys/module/amdgpu/parameters/ppfeaturemask .

Lone_Wolf wrote:

Can't prove it, but those 2 logs give me the impression that this may be wayland/xwayland specific .
Please test with a simple X environment that doesn't require gtk and/or qt.
a few options are : twm, openbox, i3 , icewm

I will check asap, I was a bit busy in the meantime

czr · 2026-09-30 10:28:59

Marichan wrote:

czr wrote:
**@Marichan** I'm still testing but looks like I've found a way to avoid this crash - approx 5h of uninterrupted gameplay with this settings:
- LACT - UV and underclock the card - In my case the sweet spot is GPU Pstate2 - 2000mhz @ 1107mV (this solved crashes during processing demanding gameplay)
- set amdgpu.ppfeaturemask=0xffff7fff in your bootloader (this avoids crashes during transitions when card downlocks and upclocks rapidly)
- capped fps at 100 (might be irrelevant)
My card is Sapphire Nitro+ RX 5700XT Black Edition. Hope this helps.
I tried full well to use LACT, this is actually how I found the reliable emthod to trigger the crash, by having the desktop showing. Undervolting, avoiding throttling and so on helps with the game and temps but I will still get the crash reliably. I also already tried the ppfeaturemask, no dice, and btw you might not haven't set it properly if you are using LACT and setting this in your bootloader. The ppfeaturemask is used to enable overdrive as well and lact will create a /etc/modprobe.d/99-amdgpu-overdrive.conf that overrides the boot option with all bits sets, so make you're not in this situation by catting /sys/module/amdgpu/parameters/ppfeaturemask .

I did set the flag after LACT did it's setup and it is set properly (thanks):

cat /sys/module/amdgpu/parameters/ppfeaturemask
0xffff7fff

Does your card have physical OC/performance switch - is OC bios selected? Maybe your GPU is power capped. I'll do more testing today and we'll see how stable it is.

Marichan · 2026-09-30 10:52:23

czr wrote:

-snip-
I did set the flag after LACT did it's setup and it is set properly (thanks):
cat /sys/module/amdgpu/parameters/ppfeaturemask
0xffff7fff
Does your card have physical OC/performance switch - is OC bios selected? Maybe your GPU is power capped. I'll do more testing today and we'll see how stable it is.

I have the OC switch, but I think it doesn't even do anything anymore on modern linux. regardless it changes nothing for the driver crash. Frankly those clocking tricks sound to me like complete placebo, if the card was having clock/power problem it would also crash with just the graphic pipeline, and running the game windowed and at a lower resolution wouldn't make any difference or at best help, while I can trigger it even more reliable in that state. To me this seems more like a synchronization issue between different queues.

I am wondering, did you try running the game you are testing windowed? If i don't and run it fullscreen the freezes are less reliable, sometime I can go an hour or so without one and sometimes it can crash in 5 minutes, but not as fast as windowedm where i get it on the titlescreen. Also do you happen to have access to monhun world? I havent tested cyberpunk because I dont have it in my library.

czr · 2026-09-30 17:21:45

Marichan wrote:

czr wrote:
-snip-
I did set the flag after LACT did it's setup and it is set properly (thanks):
cat /sys/module/amdgpu/parameters/ppfeaturemask
0xffff7fff
Does your card have physical OC/performance switch - is OC bios selected? Maybe your GPU is power capped. I'll do more testing today and we'll see how stable it is.
I have the OC switch, but I think it doesn't even do anything anymore on modern linux. regardless it changes nothing for the driver crash. Frankly those clocking tricks sound to me like complete placebo, if the card was having clock/power problem it would also crash with just the graphic pipeline, and running the game windowed and at a lower resolution wouldn't make any difference or at best help, while I can trigger it even more reliable in that state. To me this seems more like a synchronization issue between different queues.
I am wondering, did you try running the game you are testing windowed? If i don't and run it fullscreen the freezes are less reliable, sometime I can go an hour or so without one and sometimes it can crash in 5 minutes, but not as fast as windowedm where i get it on the titlescreen. Also do you happen to have access to monhun world? I havent tested cyberpunk because I dont have it in my library.

In my case physical switch increased power limit and it was visible in LACT. For me increasing frequency from 2000Mhz even by 50Hz - guaranteed HARD crash within minutes. So all this UV stuff seems to help but it's not a solution.

Yesterday I've got 5h+ of playtime without crash. Today 1-2h at most with the same settings.

Looking at logs GCVM_L2_PROTECTION_FAULT_STATUS happens every 30 minutes or so minutes in the background but the driver is able to recover without crashing desktop compositor - in game symptom is 1-2s of black screen but I can continue.

Unfortunately had 2 hard crashes today so - no fix.

WIll try RC kernel now. I'm out of ideas.

Marichan · 2026-09-30 22:36:54

Lone_Wolf wrote:

Can't prove it, but those 2 logs give me the impression that this may be wayland/xwayland specific .
Please test with a simple X environment that doesn't require gtk and/or qt.
a few options are : twm, openbox, i3 , icewm

Tried on i3, crashes all the same, in fact the crash was so bad it didn't even properly log in the journal and I had to hard reset. So I think this rules out a wayland/vulkan specific issue on the wm/compositor side

TorxScrewEnjoyer1994 · 2026-10-24 15:01:45

Hello, I registered an account to post here after reading this thread.

I experienced a year of crashes similar to yours. I believe my crashes are relevant based on your dmesg.
The crashes happen when I play CS2, and one time when playing Deadlock. It happens at random with no specific trigger that I can think of. However, I have a strong feeling that there is correlation: the game crashes more often if I play the game after waking up my computer from sleep, with Steam never having been closed.

Below are some of the previous crashes of which it left something of interest in the journal.

okt 11 20:49:49 pc kernel: amdgpu 0000:0b:00.0: Dumping IP State
okt 11 20:49:49 pc kernel: amdgpu 0000:0b:00.0: Dumping IP State Completed
okt 11 20:49:49 pc kernel: amdgpu 0000:0b:00.0: [drm] AMDGPU device coredump file has been created
okt 11 20:49:49 pc kernel: amdgpu 0000:0b:00.0: [drm] Check your /sys/class/drm/card1/device/devcoredump/data
okt 11 20:49:49 pc kernel: amdgpu 0000:0b:00.0: ring gfx_0.0.0 timeout, signaled seq=14075946, emitted seq=14075948
okt 11 20:49:49 pc kernel: amdgpu 0000:0b:00.0:  Process cs2 pid 71489 thread VKRenderThread pid 71529
okt 11 20:49:49 pc kernel: amdgpu 0000:0b:00.0: Starting gfx_0.0.0 ring reset

okt 08 21:36:33 pc kwin_wayland[1060]: atomic commit failed: Device or resource busy
okt 08 21:37:05 pc systemd[983]: Starting KRunner...
okt 08 21:37:05 pc systemd[983]: Started KRunner.
okt 08 21:37:05 pc krunner[41096]: Connecting to deprecated signal QDBusConnectionInterface::serviceOwnerChanged(QString,QString,QString)
okt 08 21:37:05 pc krunner[41096]: Failed to register with host portal QDBusError("org.freedesktop.portal.Error.Failed", "Could not register app ID: App info not found for 'org.kde.krunner'")
okt 08 21:37:06 pc kernel: amdgpu 0000:0b:00.0: Dumping IP State
okt 08 21:37:06 pc kernel: amdgpu 0000:0b:00.0: Dumping IP State Completed
okt 08 21:37:06 pc kernel: amdgpu 0000:0b:00.0: [drm] AMDGPU device coredump file has been created
okt 08 21:37:06 pc kwin_wayland[1060]: Libinput: client bug: timer button-debounce-debounce-event7: scheduled expiry is in the past (-141ms), your system is too slow
okt 08 21:37:06 pc kernel: amdgpu 0000:0b:00.0: [drm] Check your /sys/class/drm/card1/device/devcoredump/data
okt 08 21:37:06 pc kwin_wayland[1060]: Libinput: client bug: timer button-debounce-debounce-short-event7: scheduled expiry is in the past (-154ms), your system is too slow
okt 08 21:37:06 pc kernel: amdgpu 0000:0b:00.0: ring gfx_0.0.0 timeout, signaled seq=5564801, emitted seq=5564803
okt 08 21:37:06 pc kwin_wayland[1060]: Libinput: client bug: timer button-debounce-debounce-event7: scheduled expiry is in the past (-119ms), your system is too slow
okt 08 21:37:06 pc kernel: amdgpu 0000:0b:00.0:  Process cs2 pid 34731 thread VKRenderThread pid 34770
okt 08 21:37:06 pc kwin_wayland[1060]: Libinput: client bug: timer button-debounce-debounce-short-event7: scheduled expiry is in the past (-132ms), your system is too slow

okt 03 22:11:44 pc kernel: amdgpu 0000:0b:00.0: Dumping IP State
okt 03 22:11:44 pc kernel: amdgpu 0000:0b:00.0: Dumping IP State Completed
okt 03 22:11:44 pc kernel: amdgpu 0000:0b:00.0: [drm] AMDGPU device coredump file has been created
okt 03 22:11:44 pc kernel: amdgpu 0000:0b:00.0: [drm] Check your /sys/class/drm/card1/device/devcoredump/data
okt 03 22:11:44 pc kernel: amdgpu 0000:0b:00.0: ring gfx_0.0.0 timeout, signaled seq=17378653, emitted seq=17378655
okt 03 22:11:44 pc kernel: amdgpu 0000:0b:00.0:  Process cs2 pid 161082 thread VKRenderThread pid 161120
okt 03 22:11:44 pc kernel: amdgpu 0000:0b:00.0: Starting gfx_0.0.0 ring reset
okt 03 22:11:44 pc kernel: amdgpu 0000:0b:00.0: Ring gfx_0.0.0 reset failed
okt 03 22:11:44 pc kernel: amdgpu 0000:0b:00.0: GPU reset begin!. Source:  1
okt 03 22:11:44 pc kernel: amdgpu 0000:0b:00.0: BACO reset
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: GPU reset succeeded, trying to resume
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: [drm] PCIE GART of 512M enabled (table at 0x00000081FEE00000).
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: VRAM is lost due to GPU reset!
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: PSP is resuming...
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: reserve 0x900000 from 0x81fd000000 for PSP TMR
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: RAS: optional ras ta ucode is not available
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: RAP: optional rap ta ucode is not available
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: SECUREDISPLAY: optional securedisplay ta ucode is not available
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: SMU is resuming...
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: use vbios provided pptable
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: smc_dpm_info table revision(format.content): 4.5
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: SMU is resumed successfully!
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: kiq ring mec 2 pipe 1 q 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring gfx_0.0.0 uses VM inv eng 0 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring comp_1.0.0 uses VM inv eng 1 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring comp_1.1.0 uses VM inv eng 4 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring comp_1.2.0 uses VM inv eng 5 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring comp_1.3.0 uses VM inv eng 6 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring comp_1.0.1 uses VM inv eng 7 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring comp_1.1.1 uses VM inv eng 8 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring comp_1.2.1 uses VM inv eng 9 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring comp_1.3.1 uses VM inv eng 10 on hub 0
okt 03 22:11:48 pc kwin_wayland_wrapper[1147]: amdgpu: The CS has cancelled because the context is lost. This context is innocent.
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring kiq_0.2.1.0 uses VM inv eng 11 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring sdma0 uses VM inv eng 12 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring sdma1 uses VM inv eng 13 on hub 0
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring vcn_dec uses VM inv eng 0 on hub 8
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring vcn_enc0 uses VM inv eng 1 on hub 8
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring vcn_enc1 uses VM inv eng 4 on hub 8
okt 03 22:11:48 pc systemd-coredump[169147]: Process 1147 (Xwayland) of user 1000 terminated abnormally with signal 6/ABRT, processing...
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: ring jpeg_dec uses VM inv eng 5 on hub 8
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: GPU reset(1) succeeded!
okt 03 22:11:48 pc systemd[1]: Started Process Core Dump (PID 169147/UID 0).
okt 03 22:11:48 pc steam[160159]: radv/amdgpu: The CS has been cancelled because the context is lost. This context is innocent.
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: [drm] *ERROR* Failed to initialize parser -125!
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: generating OpenGL texture handle failed
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in GetSynciv(invalid call)
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
...
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: generating OpenGL texture handle failed
okt 03 22:11:48 pc kwin_wayland[1067]: Failed to allocate an offscreen texture
okt 03 22:11:48 pc kwin_wayland[1067]: 0x3: GL_CONTEXT_LOST in context lost
okt 03 22:11:48 pc kwin_wayland[1067]: generating OpenGL texture handle failed
okt 03 22:11:48 pc systemd-coredump[169149]: Process 1067 (kwin_wayland) of user 1000 terminated abnormally with signal 11/SEGV, processing...
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: [drm] device wedged, but no recovery needed
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423525] Udev event detected
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596] Udev_Event_Detail
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_subsystem:  drm
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_action:     change
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_connector:  (null)
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_devname:    /dev/dri/card1
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_devmode:    (null)
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_hotplug:    (null)
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_major:      226
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_minor:      1
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    sysname:         card1
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    syspath:         /sys/devices/pci0000:00/0000:00:03.1/0000:09:00.0/0000:0a:00.0/0000:0b:00.0/drm/card1
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    attr_name:       (null)
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423653] (dw_watch_display_connections) Time since last return from sleep = 5679776459100 ns = 5679776 ms
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423658] (recently_resumed_from_sleep_by_clocktime) Resume from sleep detected by BOOTTIME/MONOTONIC
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423662] (dw_watch_display_connections) Recently resumed from sleep detected
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423666] (dw_watch_display_connections) Pausing for 1000 millisec
okt 03 22:11:48 pc kernel: amdgpu 0000:0b:00.0: [drm] device wedged, but no recovery needed
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423525] Udev event detected
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596] Udev_Event_Detail
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_subsystem:  drm
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_action:     change
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_connector:  (null)
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_devname:    /dev/dri/card1
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_devmode:    (null)
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_hotplug:    (null)
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_major:      226
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    prop_minor:      1
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    sysname:         card1
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    syspath:         /sys/devices/pci0000:00/0000:00:03.1/0000:09:00.0/0000:0a:00.0/0000:0b:00.0/drm/card1
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423596]    attr_name:       (null)
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423653] (dw_watch_display_connections) Time since last return from sleep = 5679776459100 ns = 5679776 ms
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423658] (recently_resumed_from_sleep_by_clocktime) Resume from sleep detected by BOOTTIME/MONOTONIC
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423662] (dw_watch_display_connections) Recently resumed from sleep detected
okt 03 22:11:48 pc org_kde_powerdevil[1247]: [  1410][175410.423666] (dw_watch_display_connections) Pausing for 1000 millisec
okt 03 22:11:48 pc systemd[1]: Started Process Core Dump (PID 169149/UID 0).
okt 03 22:11:48 pc systemd-coredump[169148]: [?] Process 1147 (Xwayland) of user 1000 dumped core.
                                                  
                                                  Stack trace of thread 1155:
                                                  #0  0x00007fdc4b2256ea abort (libc.so.6 + 0x256ea)
                                                  #1  0x00007fdc481f9ae3 n/a (libgallium-26.1.7-arch1.1.so + 0xbf9ae3)
                                                  #2  0x00007fdc481fedb5 n/a (libgallium-26.1.7-arch1.1.so + 0xbfedb5)
                                                  #3  0x00007fdc47be0a61 n/a (libgallium-26.1.7-arch1.1.so + 0x5e0a61)
                                                  #4  0x00007fdc47c2b5cd n/a (libgallium-26.1.7-arch1.1.so + 0x62b5cd)
                                                  #5  0x00007fdc4b2980a2 n/a (libc.so.6 + 0x980a2)
                                                  #6  0x00007fdc4b32080c n/a (libc.so.6 + 0x12080c)
                                                  
                                                  Stack trace of thread 1147:
                                                  #0  0x00007fdc4b31e4ed syscall (libc.so.6 + 0x11e4ed)
                                                  #1  0x00007fdc47bcf98b n/a (libgallium-26.1.7-arch1.1.so + 0x5cf98b)
                                                  #2  0x00007fdc47be05e3 n/a (libgallium-26.1.7-arch1.1.so + 0x5e05e3)
                                                  #3  0x00007fdc481d5b12 n/a (libgallium-26.1.7-arch1.1.so + 0xbd5b12)
                                                  #4  0x00007fdc47e72191 n/a (libgallium-26.1.7-arch1.1.so + 0x872191)
                                                  #5  0x00007fdc47744979 n/a (libgallium-26.1.7-arch1.1.so + 0x144979)
                                                  #6  0x0000560fb94901b0 n/a (Xwayland + 0x261b0)
                                                  #7  0x0000560fb9580157 n/a (Xwayland + 0x116157)
                                                  #8  0x0000560fb9495ef6 n/a (Xwayland + 0x2bef6)
                                                  #9  0x0000560fb9496f5f n/a (Xwayland + 0x2cf5f)
                                                  #10 0x0000560fb9580679 n/a (Xwayland + 0x116679)
                                                  #11 0x0000560fb9580808 n/a (Xwayland + 0x116808)
                                                  #12 0x0000560fb951217c n/a (Xwayland + 0xa817c)
                                                  #13 0x0000560fb94879fa n/a (Xwayland + 0x1d9fa)
                                                  #14 0x00007fdc4b227781 n/a (libc.so.6 + 0x27781)
                                                  #15 0x00007fdc4b2278b9 __libc_start_main (libc.so.6 + 0x278b9)
                                                  #16 0x0000560fb94893f5 n/a (Xwayland + 0x1f3f5)
                                                  
                                                  ...

okt 01 21:27:40 pc kernel: amdgpu 0000:0b:00.0: Dumping IP State
okt 01 21:27:40 pc kernel: amdgpu 0000:0b:00.0: Dumping IP State Completed
okt 01 21:27:40 pc kernel: amdgpu 0000:0b:00.0: [drm] AMDGPU device coredump file has been created
okt 01 21:27:40 pc kernel: amdgpu 0000:0b:00.0: [drm] Check your /sys/class/drm/card1/device/devcoredump/data
okt 01 21:27:40 pc kernel: amdgpu 0000:0b:00.0: ring gfx_0.0.0 timeout, signaled seq=8988875, emitted seq=8988877
okt 01 21:27:40 pc kernel: amdgpu 0000:0b:00.0:  Process cs2 pid 64635 thread VKRenderThread pid 64674
okt 01 21:27:40 pc kernel: amdgpu 0000:0b:00.0: Starting gfx_0.0.0 ring reset
okt 01 21:27:41 pc kernel: amdgpu 0000:0b:00.0: Ring gfx_0.0.0 reset failed
okt 01 21:27:41 pc kernel: amdgpu 0000:0b:00.0: GPU reset begin!. Source:  1
okt 01 21:27:41 pc kernel: amdgpu 0000:0b:00.0: BACO reset
okt 01 21:27:42 pc kwin_wayland[1056]: Pageflip timed out! This is a bug in the amdgpu kernel driver
okt 01 21:27:42 pc kwin_wayland[1056]: Please report this at https://gitlab.freedesktop.org/drm/amd/-/issues
okt 01 21:27:42 pc kwin_wayland[1056]: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
okt 01 21:27:43 pc kwin_wayland[1056]: Pageflip timed out! This is a bug in the amdgpu kernel driver
okt 01 21:27:43 pc kwin_wayland[1056]: Please report this at https://gitlab.freedesktop.org/drm/amd/-/issues
okt 01 21:27:43 pc kwin_wayland[1056]: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'

I have many more such examples. The hardware configuration never changed. The software is usually up to date. Using a Ryzen 3700X and an MSI Radeon RX 5700 MECH OC. Not exactly your card, but I believe it is so similar that it behaves the same.

I will try RADV_QUEUE_DISABLE=compute to see if it makes a difference on my end, I have not tried it before. Please let me know if there is anything you would like me to try in quest of solving this issue.

Love, Merel

LukynZ · 2026-10-24 19:32:43

Please post here your motherboard, CPU and PSU.

RX 5700 XT crashes are mostly done by bad (cheap) motherboards. I went through this too and after testing everything problem was with motherboard and its power circuits.
It was unable to deliver stable power under a lot if circumstances and Navi10 is very sensitive chip

Last edited by LukynZ (2026-10-24 19:33:08)

TorxScrewEnjoyer1994 · 2026-10-24 20:48:45

LukynZ wrote:

Please post here your motherboard, CPU and PSU.

Motherboard: Gigabyte AB350-Gaming 3 -CF
CPU: AMD Ryzen 7 3700X
PSU: Cooler Master MWE 450 (MPE-4501-ACABW)

System is running from a single NVMe drive.

LukynZ · 2026-10-25 00:58:33

Your motherboard should have a little better power cascade, however it is still average/lower average.
PSU is probably close to limits, but should be ok if is in fine condition.

I suppose you have already tested RAM for faults.

I do not want to say change motherboard and your problems are gone. However entire internet is about RX5700XT random crashes and instabilities with same symptoms without clear solution.
My solution after years(!) of crashes and lot of reading and testing (funny how many people had similar configuration as me) was change from Gigabyte B450M-S2H to Asus TUF Gaming B550M-Plus. All problems magically disappear. The power cascade solution on Asus is much more robust and better cooled.

So if you have a chance to test another, mainly better, MB or buy/test the TUF one, this is the way I would go for stability.

Marichan · 2026-10-25 16:07:38

I don't think again power problems really explains much if the system is stable when the compute queue is disabled.
That said @TorxScrewEnjoyer1994 your problem doesn't seem related to mine completely, as I see in one of your crash that whatever you were using was using openGL, and afaik you shouldn't be able to use asynchronous compute and video queue on openGL? or maybe I am misremembering. that said you can try experimenting with the compute queue disable envar and see if it makes a difference

LukynZ · 2026-10-25 17:41:09

Do you realize, that RADV_QUEUE_DISABLE=compute create lower requirement for power of the GPU as you disable parallel compute queues. That means GPU do not eat so much power and part of GPU is idling. And also CPU can wait for GPU as you disable other computing ways. That is the reason you have lower FPS then...
This can be enough for stabilizing power circuits of your MB.

However, I do not say, that MB change is universal fix for everyone. It always needs to test other parts of your computer.
I only say, that this problems with RX 5700XT are very common, very common at boards with weaker power solutions and after all testing I came to conclusion, that it is not GPU problem or drivers problem, but it is really motherboard problem with cheaper power solution.

Last edited by LukynZ (2026-10-25 17:41:46)

Marichan · 2026-10-25 17:56:12

LukynZ wrote:

Do you realize, that RADV_QUEUE_DISABLE=compute create lower requirement for power of the GPU as you disable parallel compute queues. That means GPU do not eat so much power and part of GPU is idling. And also CPU can wait for GPU as you disable other computing ways. That is the reason you have lower FPS then...
This can be enough for stabilizing power circuits of your MB.
However, I do not say, that MB change is universal fix for everyone. It always needs to test other parts of your computer.
I only say, that this problems with RX 5700XT are very common, very common at boards with weaker power solutions and after all testing I came to conclusion, that it is not GPU problem or drivers problem, but it is really motherboard problem with cheaper power solution.

Well, I can reliably trigger the crash on a titlescreen, at a low resolution, with the card power limit set lower, by just turning on and off the compute queue, it doesn't really apply here. Considering that also most of the power that a desktop class videocard use comes straight from the psu by the external connectors and the inputs are heavily filtered and the gpu has its own power regulation I doubt the motherboard could ever explain what you are observing, but again, it is irrelevant for the problem at hand here

LukynZ · 2026-10-25 18:58:26

Do you have a Windows? Does it happens in the Windows too?
If so, can you check the Event log? Will you find there "Cache Hierarchy Error" problem?

TorxScrewEnjoyer1994 · 2026-10-26 22:08:24

In my case RADV_QUEUE_DISABLE=compute did not noticeably reduce crashing.

LukynZ wrote:

So if you have a chance to test another, mainly better, MB or buy/test the TUF one, this is the way I would go for stability.

I'd like to. Unfortunately second hand prices are as high as retail prices of three years ago. Let's see if I can get a better motherboard for tests. Until then I can only hope someone discovers a bug in the driver.

Arch Linux

#1 2026-09-24 17:31:30

AMDGPU crash with kernel freeze / instability on RX 5700 XT

#2 2026-09-25 10:51:17

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#3 2026-09-25 19:12:16

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#4 2026-09-26 11:22:47

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#5 2026-09-26 15:06:46

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#6 2026-09-27 18:46:12

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#7 2026-09-29 00:14:36

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#8 2026-09-29 21:27:54

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#9 2026-09-30 00:37:49

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#10 2026-09-30 10:28:59

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#11 2026-09-30 10:52:23

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#12 2026-09-30 17:21:45

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#13 2026-09-30 22:36:54

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#14 2026-10-24 15:01:45

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#15 2026-10-24 19:32:43

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#16 2026-10-24 20:48:45

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#17 2026-10-25 00:58:33

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#18 2026-10-25 16:07:38

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#19 2026-10-25 17:41:09

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#20 2026-10-25 17:56:12

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#21 2026-10-25 18:58:26

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

#22 2026-10-26 22:08:24

Re: AMDGPU crash with kernel freeze / instability on RX 5700 XT

Board footer