[BUG] Illegal instruction error when reading images on AMD A6-3420M processor

Status Closed — not planned
Maintainer reply ✓ Yes — ant-kurt
Activity 12 comments · opened Aug 20, 2025 · closed Jan 7, 2026
💡 Likely answer: A maintainer (ant-kurt, collaborator) responded on this thread — see the highlighted reply below.

Environment

  • Platform (select one):
  • [x] Anthropic API
  • [ ] AWS Bedrock
  • [ ] Google Vertex AI
  • [ ] Other:
  • Claude CLI version: 1.0.84
  • Operating System: openSUSE Tumbleweed 20250818, Kernel 6.16.1-1-default (64-bit)
  • Terminal: KDE Konsole

## Bug Description

Claude Code crashes with "Instrução ilegal (imagem do núcleo gravada)" / "Illegal instruction (core dumped)" when attempting to read image files using the Read tool. The crash occurs
specifically when processing images, while text files work normally.

## Steps to Reproduce

  1. Start Claude Code in terminal
  2. Execute: claude read "/path/to/any/image.png"
  3. Application immediately crashes with illegal instruction error

## Expected Behavior

Image should be read and analyzed by Claude Code, similar to how text files are processed.

## Actual Behavior

Instrução ilegal (imagem do núcleo gravada) claude
The application terminates immediately with a core dump.

## Additional Context

Hardware: AMD A6-3420M APU with Radeon HD Graphics (4 cores, from 2011)Memory: 6 GiB RAM (5.3 GiB usable)Desktop: KDE Plasma 6.4.4

Troubleshooting attempted:

  • Updated AMD microcode to revision 0x03000027
  • Reinstalled glibc and Mesa drivers
  • Tested compatibility environment variables:
  • PGI_FASTMATH_CPU=sandybridge
  • GLIBC_TUNABLES=glibc.cpu.hwcaps=-AVX2,-AVX512F
  • OPENBLAS_CORETYPE=SANDYBRIDGE

Working alternatives:

  • Text files: claude read /tmp/text_file.txt works normally
  • ImageMagick: identify -verbose image.png processes same images without issues

Suspected cause: AMD A6-3420M (2011) may lack modern CPU instructions required by Claude Code's image processing libraries.

View original on GitHub ↗

12 Comments

github-actions[bot] · 1 year ago

Found 1 possible duplicate issue:

  1. https://github.com/anthropics/claude-code/issues/5864

This issue will be automatically closed as a duplicate in 3 days.

  • If your issue is a duplicate, please close it and 👍 the existing issue instead
  • To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

ant-kurt collaborator · 1 year ago

Was this ever working for you on previous versions of Claude Code?

thiagocf84 · 1 year ago
Found 1 possible duplicate issue: 1. [[BUG] illegal hardware instruction, please compile without the AVX2 requirement #5864](https://github.com/anthropics/claude-code/issues/5864) This issue will be automatically closed as a duplicate in 3 days. If your issue is a duplicate, please close it and 👍 the existing issue instead To prevent auto-closure, add a comment or 👎 this comment 🤖 Generated with Claude Code

Regarding the duplicate detection: While #5864 has similar symptoms, it affects an Intel processor (i7-3635QM) whereas mine is AMD A6-3420M. These are related but distinct compatibility issues.

Was this ever working for you on previous versions of Claude Code?

Regarding previous versions: This is my first installation of Claude Code (version 1.0.84), so I cannot confirm if earlier versions worked on this AMD A6-3420M processor.

---

The suggested solution in #5864 about compiling with RUSTFLAGS="-C target-cpu=x86-64" for broader compatibility seems relevant to both issues.

diwu-sf · 1 year ago

A6-3420M https://www.techpowerup.com/cpu-specs/a6-3420m.c892
this CPU doesn't even have first gen AVX support, so the native binary will have to be compiled against baseline x86-64 to work

thiagocf84 · 1 year ago

Thank you @diwu-sf for confirming the CPU specifications! You're absolutely right - the A6-3420M doesn't have AVX support (not even first generation AVX), which explains the "Illegal instruction" error I'm experiencing.

This confirms that the native binary needs to be compiled against baseline x86-64 to work on older CPUs like mine. Would it be possible to provide an alternative binary compiled without AVX instructions, or perhaps make the npm installation method work as a fallback for systems that can't run the native binary?

I'm happy to test any potential solutions or provide additional system information if needed.

Thanks for looking into this!

thiagocf84 · 1 year ago

Update: I tested the npm installation method as suggested, but the "Illegal instruction" error persists even with the npm version.

Additional findings:

  • NPM version: 1.0.86 installed successfully
  • Ripgrep binary (included in package): Works normally without AVX
  • Node.js 22.15.1: Correctly recognizes A6-3420M CPU
  • Issue: Some native dependency in the npm package still uses AVX instructions

Both installation methods (native and npm) fail on CPUs without AVX support. The solution still requires all native dependencies to be compiled against baseline x86-64.

Would it be possible to provide a version with all dependencies compiled for maximum compatibility with older CPUs?

Thanks for the continued investigation!

macgyver2028 · 1 year ago

Commenting here and adding my experience and setup. Not critical, but interested if a solution comes from this.

  • Claude Code Version: v1.0.90
  • Opteron K10 family, so also without AVX
  • Ubuntu 24.04
  • Node.js v20.19.4
  • Using via ssh in terminal.
  • crash when reading images on filesystem, regardless of format
  • crash when reading image (png) from playwrite
thiagocf84 · 12 months ago

UPDATE: Discovered that Google Gemini CLI works fine on the same AMD A6-3420M processor.
Key difference: Gemini CLI runs on Node.js v22.15.1 (JavaScript runtime) while Claude Code appears to be a compiled binary with modern CPU instruction requirements.

github-actions[bot] · 8 months ago

This issue has been inactive for 30 days. If the issue is still occurring, please comment to let us know. Otherwise, this issue will be automatically closed in 30 days for housekeeping purposes.

thiagocf84 · 8 months ago

Hi there, thanks for the follow-up.

I just wanted to provide a bit more context from my side. Unfortunately, this image analysis issue is a critical blocker for my workflow, especially considering the hardware specifics I
detailed in the bug report.

Due to this, I've had to migrate my entire workflow—which was originally planned around using Claude Code for image analysis tasks—over to Gemini Pro (via Gemini CLI). I can confirm it has
been handling these tasks perfectly for over a month.

I'm sharing this information in case it's helpful for prioritizing and understanding the impact of this bug. I hope to see it resolved soon.

Best regards.

github-actions[bot] · 7 months ago

This issue has been automatically closed due to 60 days of inactivity. If you're still experiencing this issue, please open a new issue with updated information.

github-actions[bot] · 7 months ago

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.