Amazon.com Inc. is bringing its new AWS Graviton5 processors to more workloads with the launch of its new, memory-optimized R9g and R9gd instances, which are available today in multiple cloud regions.
They were revealed in a blog post by AWS Principal Developer Advocate Daniel Abib, who explained that they’re designed to power the most data-intensive applications, such as in-memory databases like Redis OSS, Memcached and Valkey. Other targeted workloads include big data analytics engines and high-throughput web services, as well as containerized microservices running on Kubernetes and Amazon EKS.
The R9g and R9gd instances are powered by the company’s next-generation Graviton5 central processing units, which were first announced by the company at its annual AWS re:Invent conference in December. We’ve already seen the CPUs debut in AWS’ general-purpose M9g and M9gd instances, followed by the compute-heavy C9g and C9gd instances that launched in June. As with those earlier offerings, Graviton5 brings significant architectural upgrades.
For instance, Graviton5 delivers up to 25% better compute performance, Abib said, as well as 35% faster web application execution and machine learning inference, and 30% faster database processing. The chips are paired with DDR5 8800 memory modules, which support 8,800 megatransfers per second, up from 5,600 MT/s in Graviton4. They also have a five-times-larger L3 cache and three-times-higher packet-processing performance than Graviton4, dramatically reducing the time workloads sit idle while waiting for the data they need.
Abib said the R9g and R9gd instances are available in 11 sizes, scaling from medium up to metal-48xl configurations powered by 192 virtual CPUs and 1,536 gibibytes of memory. At the upper end, the instances provide up to 100 gigabytes per second of network bandwidth and 72 Gbps of Amazon Elastic Block Store bandwidth, more than two-times the previous generation instances. According to Abib, they also support an Instance Bandwidth Configuration feature, which gives users the flexibility to reallocate bandwidth by up to 25% between EBS and virtual private cloud networking, so they can better match their database or caching throughput requirements.
The main difference between the two instances is that R9g relies exclusively on attached EBS volumes, whereas R9gd supports high-density local NVMe SSD storage at up to 11.4 terabytes with the largest 48xlarge sizes. By using local storage, the R9gd instances can leverage ultra-low-latency performance to support temporary caches, scratch disks and distributed analytics workloads.
Another key feature is the AWS Nitro System, which incorporates the Nitro Isolation Engine that was first introduced with the M9g and C9g instance families. NIE offloads tasks such as virtualization, storage and networking to dedicated hardware, achieving bare metal-like performance while maintaining strict workload isolation for security purposes.
Abib said the new instances are available from today in several AWS regions, including US East (N. Virginia, Ohio), US West (Oregon), and Europe (Frankfurt), with On-Demand, Savings Plans, Spot Instances, Dedicated Instances and Dedicated Host pricing options.
