g4dn.2xlarge

Analysis

The g4dn.2xlarge is a gpu instance instance in the g4dn series with 8 vCPUs and 32 GiB of memory on Intel Xeon Family (up to 2.5 GHz), which makes it balanced at 4 GiB of memory per vCPU.

Within the same generation and size, the g4ad.2xlarge on AMD EPYC 7R32 costs $0.5412/hour, 28% less ($154 a month).

It carries 1 GPU with 16 GB of GPU memory, so most of the price is the accelerator rather than the CPUs. It has local instance storage (225 GB NVMe SSD), which is fast but lost when the instance stops. Network: Up to 25 Gigabit.

Running it in N. Virginia instead of São Paulo saves $384 a month per instance: $0.7520/hour against $1.28/hour, a 70% spread.

It is size 2 of 6 in the g4dn series. Stepping up to the g4dn.4xlarge gives 16 vCPUs and 64 GiB for $330 more a month. Dropping to the g4dn.xlarge saves $165 a month. Unit price is not flat across the series ($0.0680 to $0.1315 per vCPU-hour), so the size changes the rate you pay.

On demand it costs $0.7520/hour in US East (N. Virginia), $549 a month if left running, which puts it mid-table for gpu instance types at the 38th percentile per vCPU ($0.0940 per vCPU-hour) and the 53rd percentile per GiB ($0.0235 per GiB-hour).