Last week Amazon/AWS announced the new EC2 M8a instances as their latest-generation, general-purpose compute instances now powered by AMD EPYC 9005 “Turin” processors. Amazon announced the M8a as having up to 30% higher performance and up to 19% better price performance over M7a. With my testing of both at 32 vCPUs, the new AMD EPYC Turin instance provided 1.59x the performance over the prior-generation EPYC Genoa instance!
The Amazon EC2 M8a instances are now generally available for offering 5th Gen AMD EPYC in Amazon’s cloud for general purpose VMs. Amazon Web Services is promoting the new M8a instances as great for in-memory databases, distributed caches, real-time analytics, and general purpose computing. Each M8a vCPU is backed by a physical Zen 5 CPU core. With the upgrade from AMD EPYC 9004 to EPYC 9005 or M7a to M8a is also the DDR5-4800 to DDR5-6400 memory upgrade for much greater memory bandwidth.
As I’ve shown over the past year in many Phoronix articles of AMD EPYC 9005 benchmarking, these Zen 5 EPYC processors offer incredible performance and power efficiency. As my first rodeo of EPYC Zen 5 in the cloud, I ran some benchmarks of Amazon EC2 M7a and M8a instances for looking at the performance and value of a generational upgrade.
The M8a instance was powered by custom AMD EPYC 9R45 processors. Again, Amazon Web Services is deploying these M8a instances with DDR5-6400 memory.
As I was testing these M7a/M8a instances on my own without involvement from AWS or AMD, I tested just the 32 vCPU instances in order to keep costs low. The m8a.8xlarge and m7a.8xlarge instances were tested while using Ubuntu 24.04 LTS with the Linux 6.14 kernel and GCC 13.3 compiler. Both instances were similarly configured for a 1:1 generational comparison from M7a to M8a.
The on-demand costs $1.85472 USD per hour for the m7a.8xlarge and $1.94752 USD per hour for the new m8a.8xlarge.
