g7.4xlarge

Analysis

The g7.4xlarge is a gpu instance instance in the g7 series with 16 vCPUs and 64 GiB of memory on Intel Xeon Scalable (Granite Rapids) (up to 3.9 GHz), which makes it balanced at 4 GiB of memory per vCPU.

Compared with the previous generation, it costs 130% more than the g6.4xlarge on AMD EPYC 7R13 ($1,255 more a month per instance), in exchange for the newer processor.

Running it in N. Virginia instead of Frankfurt saves $556 a month per instance: $3.04/hour against $3.80/hour, a 25% spread. It is not offered in Canada Central, São Paulo, Ireland, London, Mumbai, Singapore, Sydney and Tokyo, so plan around that if you need a presence there.

It carries 1 GPU with 32 GB of GPU memory, so most of the price is the accelerator rather than the CPUs. It has local instance storage (1x875 GB NVMe SSD), which is fast but lost when the instance stops. Network: Up to 100 Gigabit.

It is the cheapest of the 2 processor variants AWS sells at this size and generation. The g7e.4xlarge on Intel Xeon Scalable (Emerald Rapids) costs 31% more, and no others also sit above it.

It is size 2 of 6 in the g7 series. Stepping up to the g7.8xlarge gives 32 vCPUs and 128 GiB for $762 more a month. Dropping to the g7.2xlarge saves $381 a month. Unit price is not flat across the series ($0.1277 to $0.3150 per vCPU-hour), so the size changes the rate you pay.