Amazon Elastic Compute Cloud R8g instances featuring the high-performance Graviton4 chips are now available for general use on AWS. These instances, which have been in preview since last November, focus on energy efficiency, power, and performance. They can be accessed in the AWS East US (N. Virginia), East US (Ohio), West US (Oregon), and Europe (Frankfurt) Regions.
The Graviton4 processor delivers up to 30% better performance compared to the previous Graviton3-based Amazon EC2 R7g instances, making it ideal for memory-intensive workloads like high-performance databases and real-time big data analytics. In addition to improved performance, the R8g instances offer a choice of cloud processors, including Intel and AMD’s x86 architecture.
Key improvements in the R8g instances with Graviton4 include better performance for web applications, databases, and large Java applications, up to 3x more vCPUs, 3x more memory, 75% more memory bandwidth, 2x more L2 cache, and increased network and EBS bandwidth. These instances take advantage of offloaded CPU virtualization, storage, and networking features for enhanced performance and security.
For the first time, Graviton instances offer two bare metal sizes (metal-24xl and metal-48xl) on R8g instances. AWS has emphasized the energy efficiency of Graviton4, stating that it is the most energy-efficient processor they have designed. By offloading CPU, storage, and network virtualization functions, AWS aims to reduce energy consumption and carbon footprint by up to 60% compared to other Amazon EC2 instances.
Organizations considering Amazon EC2 R8g instances for hosting high-performance workloads in the cloud have a range of options to choose from. In addition to Graviton processors, they can explore other Amazon cloud computing services or infrastructure-as-a-service providers.
Overall, the availability of Amazon EC2 R8g instances with Graviton4 chips provides users with improved performance, energy efficiency, and a variety of options for optimizing their cloud workloads.
Article Source
https://www.techrepublic.com/article/aws-ai-training-chips-available/