Open-source bring-up + verified matmul on the first-gen AMD XDNA1 (Phoenix/Hawk Point) NPU on Linux — the gen FastFlowLM/Lemonade skip. RyzenAI-npu1, mlir-aie/IRON, XRT.

Native ROCm acceleration for Ollama on the AMD Radeon 780M iGPU (gfx1103) on Ubuntu. One-command setup that builds ollama-for-amd from source, applies three small ml/device.go patches, and installs Fedora 43's prebuilt gfx1103 Tensile kernels into the system rocBLAS.

llama.cpp fork — gfx1103 / Radeon 780M MMVQ nwarps decode fix (+22–36% tg). See GFX1103‑780M‑FIX.md

Complete AMD driver pack (GPU + Audio + NPU) for Windows Server 2025 - AMD Phoenix/Radeon 780M