AI is rapidly changing. Models become larger and more compute and memory intensive. This requires specialized architectures to provide high performance while taking little power and occupying little space. AI is also becoming pervasive with inference needs from the edge to the cloud and HPC systems. The AMD AI Engine provides AI acceleration that scales from laptops to data center with optimized compute and data movement to maximize reuse and latency tolerance. This talk will cover the range of systems targeted by the AI Engine and provide an overview of its architecture and the major features providing high performance and energy efficiency.
Alex Rico
Alex is a Principal AI Processor Architect in the AI Engine Architecture team at AMD. Previously he was a Principal Research Engineer at Arm investigating designs for future datacenter/HPC systems in the Architecture Research team. He has worked on large-scale systems since he was a Postdoctoral and Junior Researcher at the Barcelona Supercomputing Center where he contributed with new simulation techniques, scalable system designs and co-design with tasking programming paradigms. Alex has co-invented more than 15 patents and co-authored over 40 papers in the computer architecture space.
Javier Cabezas
Javier is a Principal Computer Architect in the AI Engine Architecture team at AMD, where he works on programmability and performance portability. Prior to joining AMD/Xilinx, he worked at NVIDIA as one of the main contributors implementing Unified Memory support in CUDA. His focus is on leveraging hardware/software co-design to make high-performance accelerators more accessible and easier to use. Javier earned his PhD at the Universitat Politecnica de Catalunya in Spain.
Join Zoom Meeting
