Deployment guides: local LLMs on real hardware, step by step

Strix Halo (Ryzen AI Max+ 395)

• Deploy a local LLM on Strix Halo (Ryzen AI Max+ 395): llama.cpp from zero to first answer

  The exact path we use on our own boxes: memory setup the BIOS won't tell you about, a pinned llama.cpp container on Vulkan, and the three checks that prove the server actually works — including the two failure modes that hide behind green statuses.

  Time: 45 min

• Ollama on Strix Halo (Ryzen AI Max+ 395): install, GPU detection, first model

  The default-path Ollama setup on a Strix Halo box, with the one log line that confuses everyone decoded: why the Vulkan path drops your iGPU, why ROCm picks it up anyway, and how to verify the model actually landed on the GPU instead of silently running on CPU.

  Time: 20 min

DGX Spark (GB10)

• Deploy a large MoE on DGX Spark with vLLM: single node and a two-node pair

  The deployment order that avoids the silent hangs we hit on our own GB10 pair: weights and image pinned, head before peer, readiness by endpoint not by feel, and the kernel-selection check that separates the fast configuration from the one most people run.

  Time: 2-3 h

Apple Silicon

No guide yet — honestly. The M1 Max and M4 Pro machines are declared lanes on our testbed with zero published runs. The guide will be written while deploying on them for real, not before.

Consumer GPU

No guide yet — honestly. No CUDA-lane runs published yet. When the discrete-GPU lane starts, its guide will be written from the deployment, not from other people’s posts.