g5g.16xlarge

Analysis

The g5g.16xlarge is a gpu instance instance in the g5g series with 64 vCPUs and 128 GiB of memory on AWS Graviton2 (up to 2.5 GHz), which makes it compute-heavy at 2 GiB of memory per vCPU.

Running it in N. Virginia instead of Singapore saves $800 a month per instance: $2.74/hour against $3.84/hour, a 40% spread. It is not offered in Ohio, Canada Central, São Paulo, Ireland, London, Mumbai and Sydney, so plan around that if you need a presence there.

It carries 2 GPUs with 32 GB of GPU memory, so most of the price is the accelerator rather than the CPUs. Storage is EBS only. Network: 25 Gigabit.

At $2.74/hour on demand in US East (N. Virginia), about $2,003 a month, the g5g.16xlarge is one of the cheaper ways to buy gpu instance capacity: only 1% of the 78 current gpu instance types cost less per vCPU-hour than its $0.0429.

It is the cheapest of the 2 processor variants AWS sells at this size and generation. The g5.16xlarge on AMD EPYC 7R32 costs 49% more, and no others also sit above it.

It is the largest of the 5 g5g sizes. Dropping to the g5g.8xlarge saves $1,002 a month. Unit price is not flat across the series ($0.0429 to $0.1050 per vCPU-hour), so the size changes the rate you pay.