At AMD we harness the power of our world-class silicon to push the boundaries of software innovation through internal R&D, ISV partnerships, and our broader engagement with the open-source software community. In July '23, AMD launched an Early Access program for Ryzen AI software to support Windows Studio Effects and today marks the General Availability (GA)1.
Ryzen AI software enables ONNX Runtime applications on the Ryzen™ AI neural processing unit (NPU) powered by a full suite of software tools and Vitis™ AI Execution Provider (EP). This allows developers to build and deploy models trained in PyTorch or TensorFlow and run them on select laptops powered by Ryzen™ AI, all made possible with AMD XDNA™ architecture, the first dedicated AI processing silicon on a Windows x86 processor.2 The XDNA architecture is purpose-built to run AI workloads efficiently and locally and offers a host of benefits for developers creating innovative AI applications. One of the unique features of AMD XDNA architecture is its ability to execute multiple concurrent spatially isolated AI workloads allowing users to operate multiple accelerated AMD AI applications without affecting each other’s performance3.
Ryzen AI software focuses on enhancing the ‘Ease of Use’ for a developers with a single click installation, array of developer resources and a growing pre-quantized model zoo hosted on Hugging Face which supports a wider variety of models (Yolo, Vision, etc) that are optimized for Ryzen AI. Both ONNX Runtime C++ and Python are supported, and models have been tested to work out of the box to make experimentation easy and deployment smooth. Provided are several step-by-step example designs to help developers kick start their application development. In addition to latest generative AI models, the release includes early access support for Automatic Speech Recognition (ASR) models like Whisper, enabling developers to build natural language speech interfaces for their applications and enable other Natural Language Processing (NLP) features such as document summarization, email assistance etc.
AMD is continually upgrading its compilation and quantization tools to offer improved performance and accuracy compared to earlier versions, and future Ryzen AI software releases are expected to include powerful profiling and optimization tools to enhance user productivity. AMD is also collaborating with leading ISVs to deliver a range of tools that collectively enhance productivity and creativity while protecting user privacy. In the near future, we expect that developers will be able to use larger LLMs to unlock a much wider range of applications, including personalized content recommendation, on-device personal assistants, productivity enhancement tools, and healthcare and financial software. Collectively, these will help the end-user save time and improve overall productivity.
To spark AI innovation amongst the developer community, AMD is collaborating with Hackster.io to host a Pervasive AI Developer Contest and up to 200 Ryzen AI powered PCs are being given to qualifying participants. Visit the contest page for more information on the rules, requirements, and hardware applications.
Additional Resources:
Register for the Ryzen AI Software Introductory Webinar
Learn more about Ryzen AI Software
View Ryzen AI Software Documentation
AI Developer Contest Overview and Rules
Footnotes
- As of May 2023, AMD has the first and only available dedicated AI engine on an x86 Windows processor, where 'dedicated AI engine' is defined as an AI engine that has no function other than to process AI inference models and is part of the x86 processor die. For detailed information, please check: https://www.amd.com/en/products/ryzen-ai. PHX-3.
- Ryzen™ AI technology is compatible with all AMD Ryzen™ 7040 series processors except the Ryzen™ 5 7540U and Ryzen™ 3 7440U. OEM enablement is required. Please check with your system manufacturer for feature availability prior to purchase. GD-220.
- Based on testing by AMD Engineering as of 10/24/23, on a test system configured with an AMD Ryzen 7 7840U processor with Ryzen AI enabled, Radeon 780M Graphics, 16GB RAM, and Windows 11 Pro to measure response time impact on individual AI applications when running AMD internal AI models of Depth Estimation, Eye Gaze Correction, Super Resolution, Scene Detection sequentially and at the same time. PC manufacturers may vary configurations, yielding different results. Performance may vary based on use of latest drivers. PHXP-57.
