AWS Graviton4 (R8g) Instances — 30% Better Performance, Same Price as Graviton3

AWS Graviton4, available as R8g instances, delivers significant performance improvements over the previous Graviton3 generation. AWS benchmarks show 30% better compute performance and up to 50% better ML inference throughput compared to R7g. R8g comes in sizes from r8g.medium (1 vCPU, 8 GB RAM at $0.0504/hr on-demand) all the way to r8g.48xlarge (192 vCPU, 1.5 TB RAM at $2.4192/hr).

Pricing is approximately the same as equivalent R7g instances, meaning existing Graviton3 customers can migrate for a zero-cost 30% performance uplift. The R8g architecture uses the same ARM instruction set as R7g, so existing ARM-compiled binaries run without recompilation. AWS has confirmed R8g instances are SAP HANA certified for configurations up to 768 GB RAM.

Cross-cloud comparison: r8g.2xlarge (8 vCPU, 64 GB) at $0.4032/hr competes with Azure E8s v5 ($0.504/hr), GCP n2-highmem-8 ($0.5610/hr) and OCI E5.Flex 4 OCPU 64 GB ($0.1152/hr). OCI remains 70% cheaper but Graviton4 closes the performance gap significantly.

TCOIQ Action: If you are currently running R6g or R7g instances for memory-intensive workloads — databases, caches, analytics — migrating to R8g is one of the highest-ROI changes you can make right now. Same cost, 30% more throughput. For SQL workloads, test carefully as improved branch prediction can change query execution plans.