AWS Tunes Up Graviton5 For Agentic AI, Boosts Bang For The Buck Bigtime
AI-summarised brief · reviewed before publication
Amazon Web Services has released its Graviton5 Arm server CPU, which is comprised of four CPU blocks with 48 V3 cores each, totaling 192 V3 cores. The Graviton5 chip is used in new M9g and M9gd instances, offering increased performance and memory capacity. The chip's design features a die-to-die interconnect, allowing for higher memory bandwidth and supporting the CXL 3.0 memory extension protocol. The Graviton5 provides 2.4X more performance per socket compared to its predecessor, the Graviton4, with a trade-off in performance per watt. This upgrade is significant for supporting databases and agentic AI workloads that require low latency. The M9g instances using Graviton5 offer improved performance and instance pricing compared to previous generations.
💡 Why It Matters
- · Lower per-chiplet costs and increased transistor density enable AWS to provide more powerful instances at competitive prices, boosting its appeal to customers with demanding workloads.
- · Enhanced memory capacity and bandwidth support emerging use cases like in-memory databases.