g6.12xlarge

Analysis

The g6.12xlarge is a gpu instance instance in the g6 series with 48 vCPUs and 192 GiB of memory on AMD EPYC 7R13 (up to 2.6 GHz), which makes it balanced at 4 GiB of memory per vCPU.

It is 19% cheaper than the older g5.12xlarge on AMD EPYC 7R32, so staying on the previous generation costs more as well as being slower. The newer g7.12xlarge on Intel Xeon Scalable (Granite Rapids) is 55% dearer at $7.13/hour.

Running it in N. Virginia instead of São Paulo saves $2,351 a month per instance: $4.60/hour against $7.82/hour, a 70% spread. It is not offered in Ireland and Singapore, so plan around that if you need a presence there.

It carries 4 GPUs with 96 GB of GPU memory, so most of the price is the accelerator rather than the CPUs. It has local instance storage (4 X 940 GB NVMe SSD), which is fast but lost when the instance stops. Network: 40 Gigabit.

It is the cheapest of the 2 processor variants AWS sells at this size and generation. The g6e.12xlarge on AMD EPYC 7R13 costs 128% more, and no others also sit above it.

It is size 5 of 8 in the g6 series. Stepping up to the g6.16xlarge gives 64 vCPUs and 256 GiB for $-879.50400 more a month. Dropping to the g6.8xlarge saves $1,889 a month. Unit price is not flat across the series ($0.0531 to $0.2012 per vCPU-hour), so the size changes the rate you pay.