G2 instances contain high-performance NVIDIA GPUs, each with 1,536 CUDA cores and 4 GB of GPU memory. There are two models of G2, as described in the following table:
Model |
GPUs |
vCPU |
Memory(GB) |
SSD Storage (GB) |
g2.2xlarge |
1 |
8 |
15 |
1 x 60 |
g2.8xlarge |
4 |
32 |
60 |
2 x 120 |
These models have only 4 GB of GPU memory, so they are limited in training. However, 4 GB of GPU memory is generally enough for serving the model to end users. One of the most important factors is that G2 instances are much cheaper than P2 instances, which allows us to deploy multiple servers under a load balancer for high scalability.