Amazon's New EC2 M7a AMD EPYC "Genoa" Instances Deliver Leading Performance In The AWS Cloud

The Graviton3 and Sapphire Rapids instances did carry a lead in basic Nginx HTTPS static web content serving compared to Genoa. Though for complex web applications and involving a complete LAMP stack the showing may come out differently.

Overall I was very impressed by the performance of the new M7a EC2 instances, although not really surprised when considering all of my local 4th Gen AMD EPYC benchmarking as well as conducting various generational tests with Microsoft's Azure cloud. The M7a instance was much faster than prior AMD Zen 3 powered instances as to be expected. The M7a instance also tended to be much faster than the M7i Sapphire Rapids power instances but with a few exceptions. Most interesting though was seeing how 4th Gen AMD EPYC could compete and outperform Amazon's Graviton3 class ARM server processors. For most workloads the M7a Genoa instance was much faster than similarly sized Graviton3(E) instances. But one area where the AWS AArch64 instances tended to remain comparable was in performance-per-dollar with AWS able to provide better pricing for these instances based on their in-house processors.

When taking the geometric mean of all the benchmarks I ran across these various 16xlarge instances, the m7a.16xlarge 4th Gen EPYC instance was 1.36x the speed of the Graviton3E instance.

For those wishing to deploy demanding workloads to the cloud or even simply wanting to try out 4th Gen AMD EPYC processors before making any physical server investments, the new Amazon EC2 M7a instances are very interesting and as shown by these benchmarks a terrific generational uplift and now providing leading performance on EC2.