Every crash is identical: the DX12 Present call never returns. The game
engine times out after ~17 seconds with WAIT_TIMEOUT (0x103).
System
- OS: Nobara 43 (Fedora-based), kernel 7.0.9-200.nobara.fc43.x86_64
- GPU: AMD Radeon RX 9070 XT (RADV GFX1201), 16GB VRAM
- Driver: mesa-vulkan-drivers-git-freeworld-26.2.0 / mesa-freeworld-26.1.0
- Proton: GE-Proton10-34 (Wine 10.0 Staging)
- Game: Warhammer 40,000: Darktide (AppID 1361210), DX12
- Desktop: KDE Plasma 6 / Wayland, gamescope 3.16.19
Crash signature from game console logs (100% consistent across 19 crashes)
<>
Thread main
Stall 17246 ms:
Application::update 17254 ms
RI::wait_for_fence 17246 ms
stingray::RenderInterface::wait_for_fence 0 ms
<</Thread Stall Warning Stack>>
<>
Thread renderer
Stall 17248 ms:
RI::present_frame 17248 ms
present_windows 17248 ms
present 17248 ms
<</Thread Stall Warning Stack>>
Game-side error code: Detected shutdown with error code: 0x103 (WAIT_TIMEOUT)
Crash signature from systemd coredump (Mesa abort)
The GPU hang propagates to all active OpenGL clients (Xwayland,
plasma-keyboard), which abort via Mesa's unconditional abort() on GPU
context reset:
#0 abort (libc.so.6)
#1 amdgpu_ctx_set_sw_reset_status (libgallium-26.1.0.so + 0xb4f1e0)
#2 amdgpu_cs_submit_ib (libgallium-26.1.0.so + 0xb5345f)
#3 util_queue_thread_func
VRAM at time of crash
Not a memory pressure issue — VRAM usage at crash time across all sessions:
- Local VRAM used: 4.6–4.9 GB of 16 GB available
- Non-local: 20 MB used
Attempted mitigations (none resolved)
amdgpu.runpm=0 amdgpu.aspm=0 kernel parametersRADV_DEBUG=nohizDISABLE_VULKAN_OBS_CAPTURE=1VK_ICD_FILENAMES forced to radeon_icd.x86_64.jsonAMD_DEBUG=nohiz added to session environment
Notes
- Crash occurs during frame present, not during heavy compute or loading
- GPU adapter correctly selected:
[DXGIHelper] choosing adapter: 0 (AMD Radeon RX 9070 XT (RADV GFX1201))
- Intel iGPU (RPL-S) also present in system but not selected
- Hardware Accelerated GPU Scheduler reports false under Wine/RADV
- Crashes date back to first use of the GPU (March 2026), 19 confirmed
crashes over ~2 months, all identical
- Game uses DX12 path; no DX11 fallback available
Every crash is identical: the DX12 Present call never returns. The game
engine times out after ~17 seconds with WAIT_TIMEOUT (0x103).
System
Crash signature from game console logs (100% consistent across 19 crashes)
<>
Thread main
Stall 17246 ms:
Application::update 17254 ms
RI::wait_for_fence 17246 ms
stingray::RenderInterface::wait_for_fence 0 ms
<</Thread Stall Warning Stack>>
<>
Thread renderer
Stall 17248 ms:
RI::present_frame 17248 ms
present_windows 17248 ms
present 17248 ms
<</Thread Stall Warning Stack>>
Game-side error code:
Detected shutdown with error code: 0x103(WAIT_TIMEOUT)Crash signature from systemd coredump (Mesa abort)
The GPU hang propagates to all active OpenGL clients (Xwayland,
plasma-keyboard), which abort via Mesa's unconditional abort() on GPU
context reset:
#0 abort (libc.so.6)
#1 amdgpu_ctx_set_sw_reset_status (libgallium-26.1.0.so + 0xb4f1e0)
#2 amdgpu_cs_submit_ib (libgallium-26.1.0.so + 0xb5345f)
#3 util_queue_thread_func
VRAM at time of crash
Not a memory pressure issue — VRAM usage at crash time across all sessions:
Attempted mitigations (none resolved)
amdgpu.runpm=0 amdgpu.aspm=0kernel parametersRADV_DEBUG=nohizDISABLE_VULKAN_OBS_CAPTURE=1VK_ICD_FILENAMESforced to radeon_icd.x86_64.jsonAMD_DEBUG=nohizadded to session environmentNotes
[DXGIHelper] choosing adapter: 0 (AMD Radeon RX 9070 XT (RADV GFX1201))crashes over ~2 months, all identical