Post-mortem GPU crash debugging with LLMs - AMD GPUOpenPost-mortem GPU crash debugging with LLMs - AMD GPUOpen

AMD FSR™<br>AMD FSR SDK AMD FSR Upscaling AMD FSR Frame Generation AMD FSR Ray Regeneration AMD FSR Radiance Caching AMD FidelityFX SDK Super Resolution 3 (FSR 3) Super Resolution 2 (FSR 2) Super Resolution 1 (FSR 1) AMD FidelityFX SDK v1 Blur Breadcrumbs library Brixelizer/GI Ambient Occlusion (CACAO) Contrast Adaptive Sharpening (CAS) Denoiser Depth of Field (DoF) Lens HDR Mapper (LPM) Parallel Sort Downsampler (SPD) Screen Space Reflections (SSSR) Variable Shading TressFX Developer testimonials FidelityFX Naming Guidelines

Cauldron Framework AMD FSR Naming Guidelines<br>Hybrid Shadows Hybrid Stochastic Reflections

Tools<br>What Tools Do We Have? Radeon Developer Tool Suite Radeon™ GPU Detective Radeon™ Raytracing Analyzer Radeon™ GPU Profiler Radeon™ GPU Analyzer Radeon™ Memory Visualizer Radeon™ Developer Panel

GPU Reshape Compressonator Frame Latency Meter OCAT

SDKs<br>What SDKs Do We Have? AMD Radeon Anti-Lag 2 SDK AMD Driver/Hardware SDKs AMD GPU Services AMD Device Library eXtra Advanced Media Framework Streaming SDK GPU Performance API

Content Creation Radeon™ ProRender Suite Radeon™ ProRender SDK GPUOpen MaterialX Library Radeon™ Rays

Vulkan® Memory Allocator Direct3D®12 Memory Allocator HIP Ray Tracing Orochi Capsaicin Framework (GI-1.0) Render Pipeline Shaders Brotli-G SDK Dense Geometry Format SDK

Platform Support<br>Unreal Engine FSR Unreal Engine 5 plugin Unreal Engine Performance Guide AMD Schola (Unreal NPCs)<br>Unity Unity CPU Profiling Guide<br>Vulkan® Radeon™ Vulkan® Drivers Version Table Developing Vulkan® Applications<br>DirectX®12 DirectX®12 Ultimate Developing DirectX®12 Applications

Docs/Research<br>Meet all our blogs Getting Started Getting Started with our Software Getting Started with Development How to Become a Graphics Programmer General Developer Tech Articles

Popular articles Integrating Anti-Lag 2 SDK Matrix Compendium Mesh Shaders Work Graphs Crash Course in Deep Learning (Graphics)

Our Publications Advanced Rendering Research Group AMD Lab Notes (HPC)<br>AMD RDNA™ Performance Guide AMD GPU Architecture Machine-readable ISAs AMD Ryzen™ Performance Guide CPU Performance and Optimization<br>Software Manuals Presentations Samples

News/Events<br>Latest Developer News Recent Software Releases Hot new articles AMD FSR 'Redstone' SDK + the neural rendering future AMD FSR Upscaling 4.1 RDNA 3 support now available Latest Radeon Developer Tool Suite update AMD Schola v2: Next-gen RL for Unreal Engine Generative AI model for GI effects Neural Networks for Geometric Representation Generative AI on AMD Radeon GPUs

Events All Events GDC Digital Dragons Other events

Watch Our Videos

Share on Bluesky Share on Mastadon Share on LinkedIn Share on Twitter/X Share on Reddit Share on Facebook Share on Whatsapp Share via Email

Post-mortem GPU crash debugging with LLMs

Originally posted: July 14, 2026

Amit Ben-Moshe

Amit Mulay

Introduction

Software development is in the middle of a fundamental shift. More and more developers rely on LLM-powered coding agents — GitHub Copilot, Claude, Codex, and others — as first-class members of their workflow, not just as autocomplete engines. These tools can read, reason about, and generate source code at a level that makes them genuine force multipliers for complex engineering tasks.

GPU crash debugging has always been one of those tasks where that force multiplication is most needed. A GPU crash typically manifests as a device reset, a black screen, or a TDR (Timeout Detection and Recovery), leaving the developer with a crash dump and little else. Correlating low-level GPU state with high-level application code is painstaking work, requiring expertise in both graphics API semantics and AMD hardware internals.

LLMs, on their own, do an OK job here. They understand the vocabulary — page faults, virtual address timelines, shader disassembly, resource lifecycles — but without structured access to the crash data, they can only speculate. With the right tooling, however, the picture changes dramatically. Give an LLM direct, structured access to the AMD GPU crash dump through a purpose-built interface, and point it at the application’s source code, and it can identify the root cause in minutes and propose a fix in high-level source code, all without debug information.

This article introduces the AMD Radeon™ GPU Detective (RGD) MCP Server : an open-source tool that connects LLMs to AMD’s crash analysis pipeline, enabling automatic post-mortem GPU crash debugging.

Overview: From a single prompt to a fix in source code

The workflow is intentionally simple from the developer’s perspective. You open your application’s source code workspace in VS Code or Claude Code CLI, attach the AMD RGD MCP Server, and send one message:

“Analyze the crash dump at crash_dumps/RenderBench_DX12-20260324-131311177.rgd”

That’s it. From that single prompt, the LLM takes over, autonomously...