Stop Crashing and Start Cooking with vLLM on AMD and Lemonade Server
Last Updated on June 25, 2026 by Editorial Team
Author(s): Cody Sandahl
Originally published on Towards AI.
How I Fixed vLLM on Strix Halo and Got 3x Better Batch Throughput with Qwen3.5
I used to wonder why “curiosity killed the cat,” but now I know that those curious cats probably forgot to eat while running experiments with AI on their local machine. I can relate. This particular journey started when one of my company’s AI developers asked me how my local AI machine would handle processing 500,000 medium-complexity data classifications. Did I technically need to answer his question? No. But that darn curious cat got me and suddenly it was 1am.
After the intro, the author explains who this setup is for (ROCm-capable AMD systems needing throughput for multi-user or batch processing) and summarizes the key fixes: update to a recent Lemonade Server, override vLLM’s default GPU memory utilization (to avoid startup crashes), reduce ctx_size for safer KV cache behavior, and cap max-num-seq to prevent performance regression under higher concurrency. They also cover how vLLM fits into Lemonade Server versus llama.cpp, then walk through installation/model download steps. The core of the post details troubleshooting why vLLM initially fails (insufficient VRAM headroom), where and how to edit Lemonade’s recipe_options.json to pass vLLM arguments, and how the author calculated practical parameter values. Finally, they present benchmarking results comparing sequential vs simultaneous requests (including effects of different batch sizes and max-num-seqs), discuss concurrency limits and scheduler overhead, note LiteLLM’s overhead, and conclude with a verdict on when vLLM is worthwhile and what trade-offs to expect.
Read the full blog for free on Medium.
Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.
Published via Towards AI
Towards AI Academy
We Build Enterprise-Grade AI. We'll Teach You to Master It Too.
15 engineers. 100,000+ students. Towards AI Academy teaches what actually survives production.
Start free — no commitment:
→ 6-Day Agentic AI Engineering Email Guide — one practical lesson per day
→ Agents Architecture Cheatsheet — 3 years of architecture decisions in 6 pages
Our courses:
→ AI Engineering Certification — 90+ lessons from project selection to deployed product. The most comprehensive practical LLM course out there.
→ Agent Engineering Course — Hands on with production agent architectures, memory, routing, and eval frameworks — built from real enterprise engagements.
→ AI for Work — Understand, evaluate, and apply AI for complex work tasks.
Note: Article content contains the views of the contributing authors and not Towards AI.
Related posts
Recent Posts
