Amazon's Graviton5 processor will go head-to-head with Intel and AMD in the cloud

Amazon Web Services (AWS) is one of the largest cloud companies in the world, and for several years it has been developing and deploying AWS Graviton processors for web applications, analytics, databases, machine learning (ML) inference, gaming, video encoding, and more. The latest in-house AWS processor, designed and built in collaboration with Annapurna Labs on TSMC's 3nm process, the AWS Graviton5, is here.

2

The chiplet design features an impressive 192 Arm V3 cores, a 5X increase in L3 Cache, a 33% lower inter-core latency, 420 GB/s die-to-die bandwidth, PCIe Gen6 and DDR5-8800 memory support with a bandwidth of 800+ GB/sec. AWS notes that compared to Graviton4-based instances, the new Graviton5 offers up to 35% faster performance for AI inference, making it an ideal chip for the current agentic era. And when it comes to memory, it delivers the "fastest memory of any processor instances in the cloud."

Naturally, this means that Amazon's new M9g instances powered by AWS Graviton5 are outperforming previous-gen AWS instances powered by Intel Xeon "Cascade Lake" and AMD EPYC "Genoa" processors. And with that, AWS confirms that Meta is one of its largest customers, and is deploying Graviton5 "at scale" with tens of millions of CPU cores supporting the company's agentic AI push.

AWS has big plans for Graviton and is already deploying new Graviton5 instances across the United States and in parts of Europe. "While many Arm-based instances have been introduced across the industry, no one comes close to the breadth and depth of the AWS Graviton footprint," the announcement reads. "After five generations of custom silicon and eight years of continuous investment, Graviton powers over 350 instance types serving more than 120,000 customers, from startups to large enterprises, a robust ISV partner ecosystem, and a broad set of managed services."