Linux diagnostics · Gaming

Linux Gaming Repair Center

Diagnose crashes, Proton failures, Vulkan issues and Linux gaming performance problems.

Local analysis. No uploads. Start with what failed, collect a small piece of evidence and test one reversible change. Logs stay in this page; reset or leaving it clears the tools.

01

Proton & Wine Doctor / Crash Investigator

A startup error, a frozen display and a system hard-lock need different evidence. Choose a symptom and answer the next question; the flow explains what each observation does and does not show.

Wine, Proton, DXVK and VKD3D-Proton: which layer is failing?
Wine
Implements Windows APIs. A prefix holds the application’s registry, files and configuration; changing or deleting it can remove settings and saves.
Proton
Steam’s game runtime combines Wine with graphics translation and integration patches. Experimental and community GE builds are different releases, not guaranteed fixes. Record the exact selected build before a controlled comparison.
DXVK
Translates supported Direct3D 8–11 paths to Vulkan. Device creation depends on the installed DXVK version, its required Vulkan features and the matching process architecture.
VKD3D-Proton
Translates Direct3D 12 to Vulkan. A working DXVK game does not establish the device features required by a different D3D12 game.

Launcher authentication, required runtime components and publisher-enabled anti-cheat support are separate startup constraints. Check the first failure and the game’s own support statements before treating a graphics warning as the cause.

Read the startup investigation guide · Investigate the D3D12 path

For a static starting point, open the guide for your failure: game startup, movement stalls / hard-locks, controller input.

02

Gaming Log Inspector

Inspect a Proton, Wine, DXVK, VKD3D-Proton or journal excerpt in this browser. Findings are clues, not a confirmed root cause. GPU, PCIe and storage evidence stay separate, including when they appear together.

Synthetic anonymized Cthulhu replay — storage evidence, no final GPU error

An anonymized acceptance case derived from Dennis’s documented Cthulhu investigation. The loaded excerpt is synthetic test data, not an original machine log.

  • Documented workstation: Ryzen 7 5800X3D, Radeon RX 9060 XT 16 GiB / Navi 44, NixOS. These identifiers do not establish rendering success.
  • Lightweight games remained stable; movement in S.T.A.L.K.E.R. 2 produced 1–2-second stalls or hard-locks. Some hard-locks left no useful final GPU message.
  • Older Arch Linux testing separately reported the GPU disappearing from the PCIe bus. That earlier observation is not inserted into this no-final-error log.
  • Moving the game from SATA to NVMe did not eliminate the problem. A 112 W GPU cap avoided some hard crashes but stalls persisted; neither observation proves a cause.
  • The SATA interface/CRC errors and link resets are a distinct I/O hypothesis. GPU/PCIe, kernel/firmware and graphics-userspace hypotheses require their own evidence.

Illustrative replay only. Timestamps are synthetic; symptoms and historical versions are supplied observations. No causal ordering, failed GPU event, confirmed root cause or fix is invented. The underlying problem remains unresolved.

Read how to separate the evidence

No final GPU message does not rule out a GPU failure. Storage resets establish storage trouble, not the cause of every freeze. Investigation guide.

03

Vulkan & Graphics Stack Inspector

Hardware identification, the bound kernel driver, Vulkan device enumeration and successful game rendering are four different observations. A 64-bit host Vulkan report also does not prove that a 32-bit game runtime can load its driver.

RADV and Intel ANV are Mesa Vulkan drivers; AMDVLK and NVIDIA use different driver paths. Check the selected ICD and runtime before forcing a driver. Wayland, X11 and Gamescope add presentation context, not an automatic explanation for a crash.

Start with a read-only graphics check

Read the PCI identity and “Kernel driver in use” separately. The “Kernel modules” field lists candidates. Then inspect Vulkan in the process architecture and container used by the game.

Read-only observation

lspci -nnk

Requirements: Requires pciutils. PCI enumeration is read-only; optional detail may require additional permission.

Keep the GPU PCI ID and “Kernel driver in use” together. “Kernel modules” lists candidates and does not establish binding.

Vulkan loader and ICD guide · 32-bit runtime guide · Hardware Explorer

04

Steam Launch Options Builder

Build a temporary diagnostic launch string from reviewed options. Choose the application API and runtime first. An option is a controlled test with requirements and side effects; it is not a general performance recommendation.

A minimal Proton logging test
PROTON_LOG=1 %command%

For a Steam game actually using Proton. It creates a diagnostic log and adds logging overhead; the exact destination depends on the runtime. Remove PROTON_LOG=1 after the test. The log may contain private paths and identifiers. The builder never changes Steam.

Collect a useful startup log

05

Performance & Frame-Time Troubleshooting

Average FPS hides brief stalls. Compare frame times during the same reproducible scene and use per-core CPU load, GPU use, VRAM and I/O as context. A single snapshot cannot establish a bottleneck or a hardware fault.

What a useful comparison records

Record the scene, resolution, render scale / FSR mode, FPS cap, VSync, Gamescope use, graphics API and Proton build. Keep those fixed while testing one setting. Lower render resolution can test GPU pressure; it cannot fix a kernel reset or prove the CPU is at fault.

Repeat first-pass stutter before clearing caches. For thermal or power suspicion, observe clocks and temperatures at the same timestamps; a low power reading during a stall may be its consequence. Movement-related stalls also require storage and memory evidence.

Shader compilation and pacing · Movement stalls

06

Gaming Knowledge Base

Focused guides with commands, interpretation, safe tests, rollback and primary references. A source review is not a reproduced compatibility result.