[BUG] Claude Desktop (Windows/MSIX): recurring "GPU process gone" crashes on hybrid GPU laptop — occurs on BOTH Intel Arc and NVIDIA; MSIX blocks every GPU workaround

Status Closed — duplicate
Maintainer reply None cached
Activity 2 comments · opened Aug 12, 2026 · closed Aug 15, 2026

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

[BUG] Claude Desktop (Windows/MSIX): recurring "GPU process gone" crashes on hybrid-GPU laptop — occurs on BOTH Intel Arc and NVIDIA; MSIX blocks every GPU workaround

Preflight

  • I have searched existing issues — related but distinct: #56805, #26302, #25801, #77857 (this report adds: crashes persist even after forcing the discrete NVIDIA GPU, ruling out "just an Intel driver problem")

What's Wrong?

Claude Desktop crashes to desktop repeatedly. Every single crash is preceded by the same log line:

[info] GPU process gone: {

followed on next start by [error] Sentry caught: {.

Crashes occur under normal chat/Claude Code UI usage. Frequency was every ~1.5–2 hours; after all mitigations below, reduced to ~1 per 1–2 days — but on 2026-08-12 it crashed twice within 5 minutes (20:47 and again 20:52, ~90 seconds after relaunch).

Environment

| Item | Value |
|---|---|
| App | Claude Desktop 1.25927.0.0 (MSIX, Claude_1.25927.0.0_x64__pzs8sxrjxfjjc) |
| OS | Windows 11, HP Omen laptop |
| GPU 0 | NVIDIA GeForce RTX 4070 Laptop GPU — driver 32.0.15.9282 |
| GPU 1 | Intel Arc Graphics (iGPU, drives display) — driver 32.0.101.8860 (2026-06-25) |
| RAM | 32 GB (13+ GB free at crash time per app's own process-memory log) |

Crash timeline (from log watcher tailing Claude's main.log)

2026-08-05 16:02  GPU process gone   (Arc, pre-mitigation)
2026-08-05 18:07  GPU process gone
2026-08-05 22:40  GPU process gone
2026-08-06 14:07  GPU process gone
2026-08-09 22:58  GPU process gone   (after Windows Graphics = High performance/NVIDIA)
2026-08-12 20:47  GPU process gone   (confirmed running on GPU 0/NVIDIA via Task Manager)
2026-08-12 20:52  GPU process gone   (90 seconds after relaunch)

What I ruled out (all verified, in order)

  1. Oversized Claude Code session files — archived 8 JSONL files (20–98 MB) out of ~\.claude\projects. Crashes continued.
  2. Corrupted install / bloated cache — full MSIX removal including 13 GB stale LocalCache in %LOCALAPPDATA%\Packages, clean reinstall from the official MSIX. Crashes continued.
  3. Stale Intel driver — Arc driver updated (32.0.101.8860). Crashes continued.
  4. Intel Arc GPU itself — forced Claude onto the NVIDIA RTX 4070 via Windows Settings → Graphics → High performance (required a reboot to take effect; verified via Task Manager GPU engine = "GPU 0 - 3D"). Crash frequency dropped but crashes CONTINUE on the NVIDIA GPU (2026-08-09, 2026-08-12 ×2).
  5. Memory pressure — app's own [process-memory] log shows ~1.4–2.1 GB app usage and 11–13 GB system free near crash times.

Why users cannot work around this (MSIX limitation)

  • --disable-gpu cannot be passed: MSIX blocks direct exe launch from WindowsApps, and no app execution alias is registered.
  • ELECTRON_EXTRA_LAUNCH_ARGS and electron-flags.conf are ignored by the MSIX packaging (as documented in #26302 / #25801).
  • There is no Hardware Acceleration toggle in Settings (requested in #56805, #77857).

So a hybrid-GPU laptop user has zero supported ways to stabilize the app.

Expected behavior

Either the GPU process crash is fixed, or users get any one of:

  1. A "Disable hardware acceleration" toggle in Settings (like Chrome, VS Code, Discord, Slack), or
  2. MSIX respecting ELECTRON_EXTRA_LAUNCH_ARGS / electron-flags.conf, or
  3. Automatic fallback to software compositing after repeated GPU process loss (Chromium already detects this).

Logs

Watcher excerpts above; full main.log excerpts and Sentry event timestamps available on request — happy to attach.

What Should Happen?

It should not crash, since fixed the driver update issue, checked no other application is running when crash happens, I am on Max paid plan and using Claude for web app development.

Error Messages/Logs

In the main ticket.

Steps to Reproduce

Working on 2 claude code session, with standard tasks, it doesnt have a rhythm when it crashes, it may crash any moment, I have tried all steps locally to fix it, even running it without MCP and connectors and still crashes.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

Claude 1.28929.0 (d1a6bc) 2026-08-11T18:58:06.000Z

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

_No response_

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗