The latest chips from AWS emphasise energy efficiency

  • Amazon Web Services debuted two newly designed chips at its re:Invent 2023 conference this week – the AWS Graviton4 and AWS Trainium2.
  • Both chips boast performance improvements compared to their predecessors, with a specific focus on energy efficiency too.
  • AWS says they can be used on a broad range of customer workloads, including ML training and generative AI applications.

Amazon Web Services (AWS) is currently hosting its re:Invent 2023 conference, where a bevy of announcements were made during the keynote on day two by CEO Adam Selipsky. Among the announcements was the debut of two new AWS-designed chips – the Graviton4 and Trainium2.

As is a significant theme during this year’s event, AWS is espousing the potential generative AI applications of both chips, with energy efficiency also being a key consideration, given the hyperscaler’s eulogising of sustainability in recent years.

“With each successive generation of chip, AWS delivers better price performance and energy efficiency, giving customers even more options—in addition to chip/instance combinations featuring the latest chips from third parties like AMD, Intel, and NVIDIA—to run virtually any application or workload on Amazon Elastic Compute Cloud (Amazon EC2),” the company highlighted in a press release.

More specifically the Graviton4 provides up to 30 percent better compute performance, along with 50 percent more cores, and 75 percent more memory bandwidth than the current generation.

As for the Trainium2, it is designed to deliver up to four times faster training than first generation chips and will be able to be deployed in EC2 UltraClusters of up to 100 000 chips. AWS says this makes it possible to train foundation models and large language models in a fraction of the time, while improving energy efficiency up to two times.

“By focusing our chip designs on real workloads that matter to customers, we’re able to deliver the most advanced cloud infrastructure to them,” noted David Brown, VP of Compute and Networking at AWS.

“Graviton4 marks the fourth generation we’ve delivered in just five years, and is the most powerful and energy efficient chip we have ever built for a broad range of workloads. And with the surge of interest in generative AI, Tranium2 will help customers train their ML models faster, at a lower cost, and with better energy efficiency,” he continued.

Both of these families of chips have yielded great benefits for AWS, with the company pointing out that it has deployed more than 150 different Graviton-powered Amazon EC2 instance types globally at scale, along with building more than 2 million Graviton processors, and it has more than 50 000 customers using Graviton-based instances.

Shifting to Trainium2 it will deliver up to 65 exaflops of compute power and give AWS customers on-demand access to supercomputer-class performance, which could be used by customers to train a 300-billion parameter LLM in weeks versus months.

All of this is being made available against the backdrop of significant improvements in terms of energy efficiency.


About Author


Related News