AWS recently announced the general availability of two new Amazon EC2 instance families featuring Graviton3 – C7g and M7g in the eu-west-2 (London) region. For organisations looking to accelerate cloud adoption and harness innovation, Graviton3 offers groundbreaking performance, cost savings, and sustainability benefits versus traditional x86 processors.

Performance Leaps Ahead of x86

Built on a cutting-edge 5nm manufacturing process, Graviton3 leverages a multi-chip architecture to dramatically boost performance over previous generations. Benchmark testing showed Graviton3 chips achieving up to 2x higher floating point performance and 3x higher cryptographic workload performance compared to the already fast Graviton2.

Graviton3-powered M7g instances offer 2x higher memory bandwidth versus M5 instances, while C7g instances provide over 4x the network bandwidth of comparable C5 instances. For real-world use cases, this makes Graviton3 ideal for workloads like high performance computing, machine learning, video encoding, gaming, and other data/network intensive applications.

Significantly Lower Costs

In addition to raw performance, Graviton3 also delivers excellent per-core value. The ARM Neoverse V1 cores in Graviton3 provide over 2x greater performance per core than Intel Ice Lake processors. This directly translates into lower costs for customers when running workloads on Graviton3 instances.

Find out more about the new ARM Neoverse V2 cores which is integrated into the new Graviton4 chipsets, announced at re:Invent 2023: https://community.aws/content/2YxIMVQ4N342OeaPMYYC2ijHuQq/aws-graviton-processors

AWS estimates M7g instances provide a 4x better price-performance ratio over M5 instances for memory-bound workloads. Network-bound applications running on C7g could see cost reductions of up to 60% compared to C5. As more mission-critical workloads shift to Graviton3 to reap these savings, overall cloud spends will decline dramatically.

Leading the Sustainability Charge

With cloud infrastructure accounting for 1-1.5% of global electricity usage, the environmental impacts of data centres are coming under greater scrutiny. Graviton3 sets a new bar for energy efficiency, requiring up to 60% lower power for the same performance versus x86.

C7g instances deliver about 40% better performance per watt than Intel Ice Lake chips. For memory-driven workloads, M7g instances show nearly 50% improvement in performance per watt. As organisations aim to reduce their carbon footprint, Graviton3 provides the most sustainable option for cloud computing today.

Compared to Intel Xeon ‘Ice Lake’-based instances, AWS Graviton3-based instance perform from 41% to 91% higher, with 39% to 71% higher performance-per-dollar, depending on instance size.

https://community.arm.com/arm-community-blogs/b/infrastructure-solutions-blog/posts/nginx-performance-on-graviton-3

Well Architected Framework (Sustainability): https://docs.aws.amazon.com/wellarchitected/latest/performance-efficiency-pillar/welcome.html

Faster Innovation for Cutting-Edge Applications

The breakthrough capabilities of Graviton3 will also accelerate deployment of innovative technologies relying on the cloud.

“Migrating to AWS Graviton3 processors has helped us save costs on scaling while empowering us to offer our users a smoother and faster experience” – Zach Pendleton, Chief Architect at Instructure

https://aws.amazon.com/solutions/case-studies/instructure-case-study/

Other emerging workloads like multiplayer cloud gaming, IoT data analytics, and video streaming will similarly benefit. Graviton3’s advantages in memory, network, and floating point performance map perfectly to these data and computation-heavy applications.

By launching Graviton3-optimised instance families like M7g and C7g, AWS is encouraging rapid migration to benefit from the next generation of cloud computing. Companies that quickly adopt Graviton3 will gain a competitive edge while sending a powerful signal to partners and customers about their commitment to technology leadership.

Adam Scott avatar

Published by