bare-metal ai stack for amd strix halo — 236 tok/s, 34 services, 4 LLMs simultaneous, 17 agents, compiled from source, zero cloud
ai amd arch-linux self-hosted autonomous gaia watchdog bare-metal homelab lemonade rocm ai-agents unified-memory no-cloud llm llama-cpp strix-halo gfx1151 compiled-from-source
-
Updated
Apr 5, 2026 - Python