The Crew Motorfest AMD GPU Crash Fix

Support
1

The Crew Motorfest on Linux: GPU crash on AMD cards – guide and report

By SAGOIST · October 4, 2026

Summary

The Crew Motorfest crashed on a Radeon RX 7900 XTX under Linux after a few minutes of driving and took the whole graphics card down with it. After switching from GE-Proton to proton-cachyos 11.0-20260702, the game ran for almost two hours without a single GPU error.

  • Problem: four crashes in 35 minutes, each with a full reset of the graphics card. Desktop programs such as the browser and Discord crashed along with the game.
  • Fix: in Lutris, switch the game’s runner to proton-cachyos.
  • Not the cause: heat, RAM or VRAM. Changing the kernel and Mesa made no difference either.
  • Confidence: one system, one day of testing. The exact cause inside GE-Proton has not been proven.

Guide: Am I affected?

You are affected if, after a crash, the kernel log shows ring gfx_0.0.0 timeout together with the process TheCrewMotorfes. This applies to AMD cards on the open-source driver (amdgpu and RADV).

Typical symptoms:

  • The picture freezes after a few minutes of driving, then the screen goes black briefly.
  • It is not only the game that dies: the browser, Discord and other programs crash too.
  • The desktop usually comes back, but is often damaged enough that a reboot is needed.

To check after a crash, run this in a terminal:

journalctl -k -b 0 | grep -Ei 'page fault|ring .* timeout|GPU reset|MES failed|VRAM is lost'

If you rebooted after the crash, replace -b 0 with -b -1. That shows the log of the previous boot.

If you see lines with TheCrewMotorfes, ring gfx_0.0.0 timeout and GPU reset, it is the same failure as in this report. If the output is empty, you have a different problem and this guide probably will not help.

Guide: The fix, step by step

Switch the game’s runner in Lutris from GE-Proton to proton-cachyos. The tested version is cachyos-11.0-20260702-slr.

  1. Download the version from the proton-cachyos release page. Take the x86_64 archive for the tag cachyos-11.0-20260702-slr.

  2. Close Lutris and Ubisoft Connect completely.

  3. Extract the archive into the Lutris runner folder:

    tar -xf proton-cachyos-*.tar.* -C ~/.local/share/lutris/runners/wine/

  4. Start Lutris, right-click The Crew Motorfest and choose “Configure”.

  5. In the “Runner options” tab, set “Wine version” to proton-cachyos-11.0-20260702-x86_64 and save.

  6. Start the game and drive for at least 30 minutes. Before the fix, the crash came after about 10 minutes at the latest.

  7. Afterwards, use the command from the section above to check that the kernel log stays empty.

Notes:

  • A new Wine prefix was not needed. The game kept running in the existing Ubisoft Connect prefix.
  • The folder name under runners/wine/ can differ slightly depending on the archive. Lutris shows exactly that folder name as the Wine version.
  • The fix has not been tested with Steam or Heroic. The matching folder there would be compatibilitytools.d.

Guide: What did not help

Three obvious measures did not stop the crashes. You can skip them.

Measure Result
Replacing GE-Proton11-6 with GE-Proton11-7 Crash with both versions
Updating the kernel from 7.2.7 to 7.2.8 Crash on both kernels
Downgrading Mesa to the previous version Crash after three minutes
Keeping Mesa 26.2.3 Crash as long as GE-Proton was active

You do not need to change cooling or graphics settings. During the stable run the card stayed below 70 °C at the hotspot, and VRAM was only a little over one third full.

Report: Test system

All data comes from a single machine, measured on October 4, 2026.

Component State
Graphics card Sapphire NITRO+ Radeon RX 7900 XTX Vapor-X, 24 GB
CPU AMD Ryzen 7 7800X3D
RAM 64 GB
Distribution Fedora 44, KDE Plasma (KWin 6.7.5), Wayland
Kernel 7.2.8-200.fc44, previously 7.2.7-200.fc44
Graphics driver amdgpu in the kernel, RADV from Mesa 26.2.3
Firmware linux-firmware and amd-gpu-firmware 20260916
Launcher Lutris with umu, Ubisoft Connect in a Wine prefix
Runners that crashed GE-Proton11-6, GE-Proton11-7
Stable runner proton-cachyos 11.0-20260702

Report: Failure pattern

Every crash followed the same sequence: the game’s render thread blocks the graphics queue, and because the targeted reset fails, the driver resets the whole card.

  1. The render thread of TheCrewMotorfes triggers a GPU page fault. The client is SQC (data), which is the loading of shader code. This entry is in the log for three of the four crashes.
  2. Two seconds later the driver reports ring gfx_0.0.0 timeout.
  3. The driver tries to reset only that queue. The GPU firmware does not answer (MES failed to respond to msg=RESET).
  4. A full reset of the card follows (MODE1 reset). All VRAM contents are lost.
  5. Every program with a GPU context crashes. In the test these were Xwayland, Discord, Brave and the Plasma on-screen keyboard.

Excerpt from the kernel log of the crash at 13:09, shortened and without the host name:

13:09:42 amdgpu 0000:03:00.0: [gfxhub] page fault (src_id:0 ring:24 vmid:7 pasid:181)
13:09:42 amdgpu 0000:03:00.0:  Process TheCrewMotorfes pid 16520 thread RenderThread pid 16585
13:09:42 amdgpu 0000:03:00.0:   in page starting at address 0x00008000030b3000 from client 10
13:09:42 amdgpu 0000:03:00.0: GCVM_L2_PROTECTION_FAULT_STATUS:0x00701430
13:09:42 amdgpu 0000:03:00.0:          Faulty UTCL2 client ID: SQC (data) (0xa)
13:09:42 amdgpu 0000:03:00.0:          PERMISSION_FAULTS: 0x3
13:09:44 amdgpu 0000:03:00.0: ring gfx_0.0.0 timeout, signaled seq=154389, emitted seq=154391
13:09:44 amdgpu 0000:03:00.0: Starting gfx_0.0.0 ring reset
13:09:46 amdgpu 0000:03:00.0: MES failed to respond to msg=RESET
13:09:46 amdgpu 0000:03:00.0: reset via MES failed and try pipe reset -110
13:09:46 amdgpu 0000:03:00.0: The CPFW hasn't support pipe reset yet.
13:09:46 amdgpu 0000:03:00.0: Ring gfx_0.0.0 reset failed
13:09:46 amdgpu 0000:03:00.0: GPU reset begin!. Source:  1
13:09:48 amdgpu 0000:03:00.0: MES failed to respond to msg=REMOVE_QUEUE
13:09:48 [drm:gfx_v11_0_hw_fini [amdgpu]] *ERROR* failed to halt cp gfx
13:09:48 amdgpu 0000:03:00.0: MODE1 reset
13:09:49 amdgpu 0000:03:00.0: GPU reset succeeded, trying to resume
13:09:49 amdgpu 0000:03:00.0: VRAM is lost due to GPU reset!
13:09:49 amdgpu 0000:03:00.0: GPU reset(1) succeeded!

In two of the crashes the faulting address was 0x0000000000000000, in one it was 0x00008000030b3000. So the address is not always the same.

Report: Timeline

All four crashes happened with GE-Proton, the only stable run with proton-cachyos. Times are CEST, and the runner for each run is confirmed by the system journal.

Launcher started Kernel Runner Result
12:22 7.2.7 GE-Proton11-6 GPU reset at 12:35, after 13 minutes
12:39 7.2.7 GE-Proton11-7 GPU reset at 12:46, after 7 minutes
12:55 7.2.8 GE-Proton11-7 GPU reset at 13:00, after 5 minutes
13:07 7.2.8 GE-Proton11-7 GPU reset at 13:09, after 2 minutes
13:21 7.2.8 proton-cachyos 11.0-20260702 no error until measurement ended at 15:20

The minutes count from the start of Ubisoft Connect, so the actual driving time was shorter.

Mesa was also changed between runs. According to the package history, Mesa was downgraded to the previous version at 13:06 and updated back to 26.2.3 at 13:11. The crash at 13:09 therefore fell into the period of the older version, the stable run into the period of 26.2.3.

Report: Measurements of the stable run

During the stable run from 13:23 to 15:20 the card stayed far below every limit. The kernel log contained no GPU error at all in that time.

Measurement Average Maximum
Hotspot temperature (junction) 64 °C 69 °C
Memory temperature 70 °C 74 °C
Card power draw 147 W 290 W
VRAM in use 8.3 GB 9.2 GB of 24 GB
Fan speed 902 rpm

Values were read once per second from the amdgpu driver’s sensors, 6,799 samples in 117 minutes. In 61% of the samples GPU load was 90% or higher.

There are no sensor values for the four crashes themselves. Measurement only started afterwards.

Limitations and open points

What is established is the link between runner and crash on one system, not the technical cause.

Established:

  • Four GPU resets with GE-Proton11-6 and 11-7, each triggered by the game’s render thread.
  • 117 minutes without an error with proton-cachyos 11.0-20260702 on an otherwise identical system.
  • Unremarkable temperatures, power draw and memory use during the stable run.

Assumed, not proven:

  • The fault lies in the translation from DirectX 12 to Vulkan (vkd3d-proton), which GE-Proton and proton-cachyos ship in different states.
  • Heat or a hardware defect can be ruled out. The stable run supports this, but sensor values from the crashes are missing.

Open:

  • A counter-test is missing: nobody switched back to GE-Proton after the stable run to trigger the crash again.
  • Only one RX 7900 XTX was tested. Whether other AMD cards or Nvidia cards are affected is unknown.
  • It is unclear whether newer GE-Proton versions fix the problem.
  • The failed queue reset that forces a reset of the whole card is a separate problem in the driver or firmware. Switching the runner only avoids the trigger.

If you have the same error, this feedback helps other players: graphics card, kernel, Mesa version, runner before and after, playtime without a crash.