g7e.12xlarge

Analysis

The g7e.12xlarge is a gpu instance instance in the g7e series with 48 vCPUs and 512 GiB of memory on Intel Xeon Scalable (Emerald Rapids) (up to 3.2 GHz), which makes it very memory-heavy at 10.67 GiB of memory per vCPU.

Running it in N. Virginia instead of London saves $4,464 a month per instance: $8.29/hour against $14.40/hour, a 74% spread. It is not offered in Canada Central, São Paulo, Ireland, Singapore and Sydney, so plan around that if you need a presence there.

Nothing else with 48 vCPUs and similar memory is cheaper: among 2 current same-shape types it has the lowest on-demand price.

It is 21% cheaper than the older g6e.12xlarge on AMD EPYC 7R13, so staying on the previous generation costs more as well as being slower.

Within the same generation and size, the g7.12xlarge on Intel Xeon Scalable (Granite Rapids) costs $7.13/hour, 14% less ($845 a month).

It carries 2 GPUs with 192 GB of GPU memory, so most of the price is the accelerator rather than the CPUs. It has local instance storage (1 x 3800 GB NVMe SSD), which is fast but lost when the instance stops. Network: 400 Gigabit.

It is size 4 of 6 in the g7e series. Stepping up to the g7e.24xlarge gives 96 vCPUs and 1024 GiB for $6,049 more a month. Dropping to the g7e.8xlarge saves $2,203 a month. Unit price is not flat across the series ($0.1646 to $0.4204 per vCPU-hour), so the size changes the rate you pay.