Native vLLM on Windows 11 for AMD RX 6000: One-Click Installer
Dizzy_Counter2481 · reddit · 2026-08-18
A developer built a native vLLM + ROCm 7.15 runtime for AMD RX 6000 series on Windows 11, featuring a one-click installer and bypassing WSL2.
Key Features:
- Native HIP/ROCm execution without WSL2 wrapper.
- Custom rocBLAS binaries for RX 6750 XT (gfx1031).
- vLLM plugin vllmwindowsrocm with TRITONATTN support.
Benchmarks (RX 6750 XT):
- rocBLAS: 25.97 TFLOPS (FP16).
- vLLM Inference: 59-62 tok/s (Qwen3.5-4B).
Implementation Details:
- Uses TheRock to build HIP/rocBLAS.
- Forces GPU recognition via HSAOVERRIDEGFXVERSION=10.3.0.
- Links PyTorch 2.12.0 against TheRock runtime.
Limitations:
- Verified on RX 6750 XT; other RDNA2 cards may need adjustments.
- Requires enforceeager=True (torch.compile disabled).
- FP8/AWQ and multi-GPU are untested.
More from coding & agent
- Open Source End-to-End Stock Report System Using LSTM and Agentic AI — kmeanskaran · 2026-08-18
- poolside teases Model Factory updates: experiment API, trajectory store, new UI — dhruv2038 · 2026-08-18
- Claude writes a MacOS driver for an obscure Windows-only HP printer — fofrAI · 2026-08-18
- Building an AI with a brain you own using Markdown and 5 tools — dfinke · 2026-08-18
- 2026 Backend Engineer Skillmap: Highlights MCP & Agents — _jaydeepkarale · 2026-08-18
- Tako-MCP: Model Context Protocol Server for Octopus Energy Japan API — modelcontextprotocol · 2026-08-18