The Evolution of AWS Graviton: Reviewing the Capabilities of Gen 1 vs Gen 2 vs Gen 3
Since launching in 2018, AWS Graviton processors continue rapid innovation across successive generations. Each new Graviton chip iteration achieves major leaps forward in performance, efficiency and capabilities.
Let’s examine the improvements as AWS maneuvers towards data center superiority over x86 servers:
Graviton Gen 1 – Laying the Foundation
The first Graviton generation focused on establishing Arm servers as viable for scale-out cloud workloads:
– Based on Arm Neoverse N1 cores
– Up to 16 cores per processor
– 45% better price/performance vs comparable x86
– Support for containers and Java/Node JS optimizations
While promising, Gen 1 lacked wins in absolute performance necessary for widespread adoption.
Graviton2 – Raising the Bar
The 2020 Graviton2 generation closed this gap by significant performance gains:
– Up to 64 next-gen Arm Neoverse cores
– 2X floating point and 4X SIMD vector processing
– 7nm manufacturing for density and efficiency
– 30% Price/Performance advantage over x86
Graviton3 – Pushing New Frontiers
Announced in 2021, Graviton3 promises a giant leap over x86 with breakthrough acceleration:
– Dual socket designs for scaling up to 128 cores
– Adds 1 to 4 tensor processor units for ML tasks
– 100Gbps network interface for ultra low latency
– Hardware root of trust for heightened security
– 2X Price/Performance projected over equal x86
In just three generations, AWS Graviton caught up with x86 performance while maintaining considerable efficiency advantages. Future Graviton iterations will likely extend leads further across video encoding, databases, analytics and machine learning – allowing more workloads to benefit from optimized architecture.
The rapid pace of innovation places AWS years ahead in tailored infrastructure capabilities. As more apps migrate to leverage Graviton acceleration, achieving infrastructure cost and capability goals grows increasingly viable for modern development teams.