AMD EPYC Turin vs. Intel Xeon 6 Granite Rapids vs. Graviton4 Benchmarks With AWS M8 Instances
First up was the QuantLib quantitative finance software where the AMD EPYC Turin instance enjoyed a hefty lead over the competition. QuantLib works on AArch64 but isn't as well optimized as on x86_64. The M8a performance was nearly twice as fast as the M8i Granite Rapids instance. As a reminder, AWS uses the physical core with HT enabled for M8i while M8a is just the physical cores without SMT.
Even with the higher price per hour of M8a, it offered the best value by far for this quantitative finance workload.
With the Incompact3d finite-difference flow solver for HPC, it performs really well on Graviton4 and allowed the AWS in-house processor to pick-up its first lead. The AMD EPYC Turin instance was right behind Graviton4 while Granite Rapids was far behind in both raw performance and value.
With the GROMACS molecular biology software, the AMD EPYC Turin powered M8a instance returned to being the front-runner with huge leads over Granite Rapids and then Graviton4 coming in last here.
NAMD also came out with significantly better performance and value with EPYC Turin compared to Intel Xeon 6 Granite Rapids. No Graviton4 results here due to relying on the official NAMD 3.0 binaries and those being x86_64 only.
Wrapping up the HPC benchmarks as part of this comparison, 5th Gen AMD EPYC continued delivering the best performance and value of these Amazon Web Services M8 instance types at 16 vCPU.