Amazon Web Services on Tuesday announced new EC2 instances powered by new versions of its custom-designed chips, the Graviton3 and Trainium.
AWS is focused on "making the full power of machine learning available for all customers," AWS CEO Adam Selipsky said during his AWS re:Invent keynote address. "Lowering the cost of training and inference are major steps of the journey."
After launching the first AWS-designed Graviton processor in 2018, the cloud company is now debuting Graviton3 for compute-intensive workloads like high-performance computing, batch processing, electronic design automation (EDA), media encoding, scientific modeling, ad serving, distributed analytics, and CPU-based machine learning inferencing.
Compared to Graviton2, the new chip will deliver up to 25% more compute performance and up to twice as much floating-point & cryptographic performance. For machine learning, Graviton3 includes support for bfloat16 data and will be able to deliver up to 3x better performance.
AWS teased the upcoming C7g instances powered by Graviton3 processors.
Selipsky also announced the new Trn1 instance, powered by Trainium. AWS last year launched Trainium, a chip purpose-built for training deep learning models. The new instances will deliver the best price-performance for deep learning training for applications like image recognition, natural language processing and fraud detection.
The instances will offer up to 800Gbps networking bandwidth. Trn1 instances are now available in preview.