Amazon has stepped up for its custom chips. Annapurna Labs which was acquired by Amazon in 2015 is a microelectronics company that has helped amazon develop its own custom chips. These chips deliver high power and provide efficient solutions. Amazon had developed its Nitro card systems in 2017 which were secure and efficient. It provided Hypervisor, network, storage, and security features. Then AWS came up with AWS Inferencia which provides Machine Learning acceleration capabilities. Finally, now amazon has released AWS Graviton2 which is its most powerful and efficient processor which is now powering EC2 instances. AWS Graviton2 processor is designed by AWS utilizing Arm Neoverse N1 cores. It is manufactured using a 7nm technology with a transistor count of 30 Billion. It is Arm v8.2 compliant with 64 cores. It has a large 64KB L1 caches and 1MB L2 cache/vCPU with Dual-SIMD units with Instructions to accelerate ML inference int8, fp16, and 8x DDR4 3200 channels with no LUMA concerns. Every core sees the same path to memory and other cores. The point to the above numbers says in English that the Graviton2 based EC2 instances provide 4x compute cores, 7x performance, and 5x faster memory vs its 1st Generation. The EC2 instances which are Graviton2 powered are M6g, R6g, C6g, T4g, C6gn, and some others. The main benefit of using this is 40% price/performance vs comparable instances. Many AWS services are now available using this including Amazon ElastiCache, Amazon EMR, Amazon RDS, and many more coming. Graviton2 also provides CPU power efficiency with higher density, lower costs, and a lower carbon footprint and increases CPU efficiency by 3x. In software support, many Linux distributions now support graviton2 and have support in docker and Kubernetes.
For further actions, you may consider blocking this person and/or reporting abuse
Top comments (2)
Great to see your interest in Graviton2, keep it up
Thanks, Jason! I am learning about Graviton2 and finding it really interesting.