m5n.8xlarge

Analysis

The m5n.8xlarge is a general purpose instance in the m5n series with 32 vCPUs and 128 GiB of memory on Intel Xeon Platinum 8259 (Cascade Lake) (up to 3.1 GHz), which makes it balanced at 4 GiB of memory per vCPU.

Within the same generation and size, the m5a.8xlarge on AMD EPYC 7571 costs $1.38/hour, 28% less ($385 a month). The other variants are the m5.8xlarge at $1.54, the m5ad.8xlarge at $1.65 and the m5d.8xlarge at $1.81.

Running it in N. Virginia instead of Tokyo saves $397 a month per instance: $1.90/hour against $2.45/hour, a 29% spread. It is not offered in Canada Central, São Paulo, London, Mumbai and Sydney, so plan around that if you need a presence there.

On demand it costs $1.90/hour in US East (N. Virginia), $1,390 a month if left running, which puts it mid-table for general purpose types at the 59th percentile per vCPU ($0.0595 per vCPU-hour) and the 59th percentile per GiB ($0.0149 per GiB-hour).

Looking beyond its own line, 21 of 37 current types with 32 vCPUs and similar memory are cheaper. The cheapest of those is the m6g.8xlarge on AWS Graviton2 at $1.23/hour.

It is size 5 of 8 in the m5n series. Stepping up to the m5n.12xlarge gives 48 vCPUs and 192 GiB for $695 more a month. Dropping to the m5n.4xlarge saves $695 a month.