AI & Compute

Recommended llama.cpp Setup for Strix Halo

The starting point we recommend on AMD Strix Halo (Ryzen AI Max+ / Radeon 8060S) for llama.cpp: ROCm as the default backend, Vulkan as the fallback, and when MTP speculative decoding is worth enabling.

New to llama.cpp on Strix Halo? This page gives the recommendation in one place; the linked pages carry the full setup steps and the numbers behind it.

The decision table

If you are…UseWhy (measured on this machine)
Running dense models day to day (e.g. Qwen3.8-27B)ROCm + MTP~106% faster generation with MTP than the ROCm baseline; +74% for Vulkan’s MTP — ROCm wins on the top end too
Running MoE models at plain decode (e.g. Qwen3.6-35B-A3B)VulkanVulkan’s baseline generation was ~14% faster than ROCm’s before any speculation
Running MoE models with MTP enabledEither — effectively a tieBoth backends converge to ~60 tok/s with MTP (ROCm 60.9 vs Vulkan 60.4)
Installing is fighting you (older Mesa, ROCm packaging issues)Vulkan / RADV from the Kisak PPAFewer moving parts; verified working on this exact hardware
Reproducing the numbers on this siteBoth builds, same llama.cpp commitThe benchmark runs both backends from one checkout (c060ca974) so only the backend differs

All figures come from ROCm vs Vulkan: llama.cpp on AMD Strix Halo — a published page with the raw JSON data linked for inspection.

Start here, in order

  1. Install ROCm 7.14 on Strix Halo — userspace-only install that keeps the kernel amdgpu driver and display working. Includes the compiler checks that prove the toolchain targets the right GPU.
  2. Vulkan / RADV compute setup — the fallback backend, from the Kisak Mesa PPA.
  3. Run llama.cpp on Strix Halo — build both backends, select the GPU explicitly, and read the practical notes: memory behavior, the MTP flag set that was used, and common failures.
  4. RAM, GTT and GPU memory on Strix Halo — what the 128 GiB unified memory actually is, what the BIOS “VRAM” number means, and how to watch GTT usage while a model is loaded.

Scope and caveats

  • These recommendations reflect one machine (Ryzen AI Max+ 395 with 128 GiB LPDDR5), two Q8_0 models, and one llama.cpp commit (c060ca974, build 10603) tested in August 2026. Upstream moves fast — after a major llama.cpp, ROCm, or Mesa upgrade, re-test before trusting old numbers.
  • The MTP win is specific to models with a matching draft head (Qwen3.x here). Other model families may not have one; the plain-decode ranking still stands as the general guide.
  • Power profile was Balanced throughout testing. A different profile changes absolute throughput, and possibly the ranking at the top end.

Sources and upstream documentation