Amazon Web Services Inc. announced Thursday the general availability of two new cloud computing instances powered by its Arm-based Graviton2 processors, which it says deliver up to 40% better price/performance over its current Intel Corp. x86-based instances.

The new AWS C6g instances were announced in December and are meant for compute-intensive workloads, including central processing unit-based machine learning inference, high-performance computing, distributed analytics, scientific modeling and video encoding. As for the memory-optimized R6g instances, they’re best paired with open-source databases such as MariaDB, MySQL and PostgreSQL, as well as in-memory caches such as KeyDB, Memcached and Redis.

Amazon said the new instances offer up to seven times the performance, four times more compute cores and five times faster memory than its existing A1 instances.

They are powered by the company’s new 64-bit Arm Neoverse cores and custom-made silicon, built using a seven-nanometer manufacturing process. When compared with AWS’ first-generation Graviton processors, the new chips offer double the floating-point performance per core for high-performance computing and scientific workloads. They can also provide optimized instructions to speed up machine learning inference, plus 50% faster per core encryption performance to boost security, the company said.

Amazon said its Elastic Load Balancing, ElastiCache and Elastic Map Reduce services have achieved “superior price/performance in testing.” It plans to move those services into production on Graviton2-based instances in the coming months.

The C6g and R6g instances are both built on Amazon’s Nitro System, which is a collection of custom-designed hardware and software technologies that enable the company to deliver more flexible and secure cloud services with features such as multitenancy and private networking.

Photo: Robert Hof/SiliconANGLE