AMD Ryzen AI Software 1.7 Released For Improved Performance On NPUs, New Model Support

Phoronix
AMD released Ryzen AI Software 1.7, featuring significant NPU performance gains and support for new LLMs and Stable Diffusion models.

Summary

AMD has launched Ryzen AI Software 1.7 for Windows and Linux, designed to enhance the utilization of Ryzen AI NPUs for AI workloads like Stable Diffusion and ONNX. This update delivers substantial performance improvements and faster compile times for the CNN/Transformer compiler. Key additions include support for new large language models (LLMs) such as Qwen-2.5-14b-Instruct and Phi-4-mini-instruct, along with preview support for Sparse-LLM on the GPT-OSS-20b model and VLM Gemma-3-4b-it, plus long context support for hybrid execution LLMs. For Stable Diffusion, support for SD3.5-Turbo and Segmind-Vega 1024x1024 has been added, and performance for all supported models using the native BFP16 format sees up to a 40% improvement.

(Source:Phoronix)