The AMD Instinct MI100 GPU accelerator just got announced earlier today. The American hardware manufacturer promises this piece of high-performance hardware will help scientific research go even further. Courtesy of the fact it’s supposed to be the world’s fastest GPU of its kind; or any kind really.
Its design is based on AMD’s latest CDNA architecture which has been specifically optimized for artificial intelligence applications. Which, in essence, means training machine learning models faster than ever before. The AMD Instinct M100 accelerator is already supported by the industry’s leading computing platforms developed by Gigabyte, Hewlett-Packard, Dell, and Supermicro.
It also offers native support for ROCm 4.0 open-source software and EPYC CPUs from AMD, as expected. The company says its key focus in creating this solution was customization. As the demands of its clients are diverse enough to warrant tailor-made infrastructural solutions. Which is why it isn’t surprising to learn Microsoft’s Azure business is already implementing a variety of the latest AMD hardware as of right now.
As for wider availability of the Instinct MI100 accelerator, AMD says it will start fulfilling enterprise orders at some point in the first quarter of 2021. Of course, a smaller number of the company’s largest customers will start receiving the first batch of the said GPU systems in a matter of days.