dmesg confirms a hardware execution unit hang on the i915 Render ring:
ecode 12:1:8ed9fff2 and ecode 12:1:85dcfdfbRender (Gen12 / RPL-P)Render context reset due to GPU hang -> Triggers VK_ERROR_DEVICE_LOST in user-space Vulkan.Also reproduces on AMD RX 9070 (RDNA4), RADV / Mesa 26.2.5, kernel 7.2.11, Proton Experimental
(and bleeding-edge). dmesg shows the same gfx ring timeout from the game's vkd3d_queue thread.
Root-caused with RADV_DEBUG=hang: the hung dispatch is compute shader 1c144c57dee5a647, which
dxil-spirv translates with inner loops that have no exit, so the overlay's culling pass spins
forever. Details, the DXIL/SPIR-V, and the hang dump: https://github.com/HansKristian-Work/dxil-spirv/issues/327
proton experimentalx2 2026-10RADV_DEBUG=hangx1 2026-10VKD3D_CONFIG=single_queue,nodxr`)x1 2026-10VKD3D_FEATURE_LEVEL=12_0x1 2026-10
Compatibility Report
System Information
Proton version: Proton Experimental [cite: 1, 5]
I confirm:
Symptoms
The game launches and regular park navigation/management renders normally on low settings. However, it consistently hangs and crashes straight to the desktop whenever the game attempts to render full-screen volumetric overlay passes:
Graphics setting adjustments (disabling Dynamic Resolution[cite: 3], setting Volumetrics to Low[cite: 4], setting Water Quality to Low[cite: 2], turning off Screen Space Reflections[cite: 2], toggling FSR/TAA[cite: 3]) and VKD3D launch configurations (
VKD3D_FEATURE_LEVEL=12_0,VKD3D_CONFIG=single_queue,nodxr) do not prevent the crash.Reproduction