I am seeing a very similar GPU reset signature on an RX 6800 (NAVI_FLOUNDER) with Mesa 26.1.5 and Linux 6.18.39.
The first fault is triggered by VS Code's Wayland/Electron GPU process, followed by:
[gfxhub] page fault
Faulty UTCL2 client ID: SQC (data)
PERMISSION_FAULTS: 0x3
ring gfx_0.0.0 timeout
Ring gfx_0.0.0 reset succeeded
device wedged, but recovered through reset
The GPU is stock, PCIe Gen4 x16, with no AER/MCE errors and normal temperatures. This is a different GPU generation and workload from this report, but the kernel fault sequence appears identical.
@JuanJoseDev88 Can you upload a save file where you can reproduce this issue?
@ooaaqq These error messages are fairly generic and look similar across different issues. Generally, if it's not from the same game, it's unlikely that it's related.
@JuanJoseDev88 Can you upload a save file where you can reproduce this issue?
@runar-work Sure here it is.
Thanks for the save file. Unfortunately, I haven't been able to reproduce this issue on my end so far. I've mainly tested on RX 6800 with 7.1.6-arch1-1, Proton Experimental and Proton GE 11-3, Mesa 26.1.5, 26.1.6, low and high graphics presets. I'm launching the game through Steam. I also tested briefly on RX 9070 and RX 7600 before I switched just in case, but no issues there either.
Do you know if any particular graphics settings trigger this problem?
BTW, if you check the "Allow using Valve Proton builds to run games" option in Heroic's advanced settings you should be able to test Proton Experimental as well. I don't know if it'll make a difference here, but it could be useful for future testing.
proton experimentalx2 2026-08ge-proton11-1x1 2026-07RADV_DEBUG=llvmx1 2026-07
FFVII Rebirth GPU Hang on AMD RX 6700 XT (RADV/Mesa) under Linux
Summary
Final Fantasy VII Rebirth consistently triggers a GPU hang on an AMD Radeon RX 6700 XT running Linux. The crash occurs reliably during the Chocobo race selection or shortly after entering a race.
The issue results in:
ring gfx_0.0.0 timeoutThe problem persists across multiple Proton versions and configuration changes.
System Information
Vulkan
The GPU supports:
PCIe
No PCIe down-training.
Overclock
GPU is completely stock.
Original Kernel Messages
The GPU consistently faults while executing the VKD3D queue thread.
Page faults:
Followed by:
Then:
Finally:
dmesg excerpt
Proton Log Symptoms
The Proton log prior to the GPU reset shows repeated VKD3D activity until the device hangs.
After the GPU reset Heroic reports the game as crashed.
Reproduction
The issue is highly reproducible.
Sometimes it crashes immediately on selection.
Sometimes it crashes 20–60 seconds later.
Testing Performed
Mesa
Tested:
No improvement.
Proton
Tested:
Behavior:
RADV_DEBUG=llvm
Tested.
Result:
ppfeaturemask
Originally:
Disabled GFXCLK DPM:
Kernel confirms:
Crash still occurs.
AMDVLK
Installed.
Forced using:
Verification:
Result:
AMDVLK therefore could not be tested further.
Heroic Logs Collected
FF7_Rebirth_crashlog.txt
The Heroic logs collected during testing consistently showed:
One crash occurred immediately after Chocobo selection.
Another occurred after several seconds inside the race.
Additional Observations
The issue appears isolated to this title.
Other Vulkan applications run normally.
GPU stress outside FFVII Rebirth has not shown instability.
The crash always involves:
and
suggesting a graphics queue hang rather than a general hardware failure.
What Was Not Tested
Suspected Cause
Current evidence suggests an interaction between:
The kernel consistently reports GPU page faults originating from the
vkd3d_queuethread, followed by graphics ring timeouts and GPU reset.No evidence currently suggests:
The behavior appears software-triggered rather than hardware-induced.
Relevant Log Snippets
Kernel
Vulkan
Proton
Requested Feedback
Has anyone observed similar GPU hangs with:
particularly during the Chocobo race sequence?
Any patches, Mesa regressions, or VKD3D workarounds would be greatly appreciated.