Amazon Web Services Taps Own Arm-Based Chips for New Supercomputing Offering

Amazon's cloud unit on Tuesday offered a new supercomputing service based on its self-designed processors, a further sign of how chips based on Arm's technology are encroaching on Intel and Advanced Micro Devices turf.

Amazon Web Services, or AWS, sells its computing services based on the customer's choice of an underlying central processor chip. Software developers have traditionally chosen between Intel or AMD products, but since 2018 Amazon has also offered its own Graviton chips designed with technology from Arm, which is in the midst of a $40 billion (roughly Rs. 2,93,600 crores) takeover by Nvidia.

Arm-based chips have long powered mobile phones because they can operate on very low power levels, but they are increasingly used in data centers where their power efficiency helps control costs. The world's fastest computing system, the Fugaku supercomputer in Japan, is based on Arm chips.

Supercomputing helps with tasks such as weather forecasting, medical research and modeling aerodynamics for cars without a wind tunnel. But systems remain expensive and mostly operated by governments and research centers.

AWS is hoping to slash costs, saying the new service will get 40 percent better price-to-performance than its similar offerings from AMD and Intel. AWS's own technology will quickly pass data through multiple Graviton processors, a key supercomputing process in which many chips act as a hive mind to tackle a large task. AWS will rent the service out so that researchers need not build or manage a system.

Supercomputing "is no longer this thing that only governments do," Dave Brown, vice president of Elastic Compute Cloud at AWS, said in an interview. "You can decrease the cost – you don't need a supercomputer any more. You can spin them up in the cloud and then spin them down."

© Thomson Reuters 2020

iPhone 12 Pro Series Is Amazing, but Why Is It So Expensive in India? We discussed this on Orbital, our weekly technology podcast, which you can subscribe to via Apple Podcasts, Google Podcasts, or RSS, download the episode, or just hit the play button below.