openai / openai/codex

ChatGPT/Codex desktop app crashes with SIGILL (illegal instruction) in renderer on AMD EPYC 7551 (Zen 1: AVX2 present, no AVX-512)

Open
#42,849 0 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

app bug Linux
Dominant language
Rust
Stars
125k
Forks
19.4k
PR merge metrics
PR metrics pending

Description

Description
The official ChatGPT desktop app with Codex for Linux crash-loops on startup/use on a Zen 1 CPU. Every crash is SIGILL (si_code: ILL_ILLOPN) in the Electron renderer process, suggesting a code path (or bundled runtime) compiled for, or JIT-emitting, instructions newer than this CPU supports. The CPU supports SSE4.2, AVX, and AVX2, but not AVX-512.

Each crash also produces a very large core dump (up to ~4.5 GB), so the crash loop fills disk quickly.

Environment

  • App: openai-codex-desktop 26.901.20858-1 (Arch Linux package; executable /usr/lib/chatgpt/ChatGPT)
  • OS: Omarchy (Arch Linux), kernel 7.1.9-arch1-2, Wayland (Hyprland), --ozone-platform=wayland
  • CPU: AMD EPYC 7551 32-Core (Zen 1 / Naples) — has AVX/AVX2/FMA/SSE4.2; no AVX-512
  • Running in a KVM/QEMU VM (Proxmox) with cpu: host passthrough — guest sees the full, real EPYC 7551 flag set

Crash evidence (systemd-coredump)

Signal: 4 (ILL) si_code: ILL_ILLOPN
Executable: /usr/lib/chatgpt/ChatGPT
Command Line: /usr/lib/chatgpt/ChatGPT --type=renderer ... --ozone-platform=wayland ...

Two identical SIGILL crashes captured on 2026-09-04 (plus earlier ones); reproducible on two separate VMs on the same host.

Expected behavior
Either run on x86-64-v3 class CPUs (AVX2 baseline, no AVX-512), or detect missing CPU features at startup and fail with a clear message instead of SIGILL, or provide a fallback build.

Steps to reproduce

  1. Install the app on a machine with a Zen 1 (or any AVX2-but-no-AVX-512) CPU.
  2. Launch it and sign in / open a chat.
  3. Renderer process dies with SIGILL; app crash-loops.

Happy to provide the core dump backtrace or run a debug build if useful.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reproducing the crash on a Zen 1 or AVX2-but-no-AVX-512 system and inspect /usr/lib/chatgpt/ChatGPT with the systemd-coredump evidence. The issue mentions a core dump backtrace and a debug build as useful next steps. Done means the app avoids SIGILL, detects unsupported CPU features with a clear message, or provides a compatible fallback.

Written by the indexing model from the issue text.

Assessment

Tech stack
electron, linux
Domain
desktop-dev, operating-systems
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.