A2 VMs Now Generally Available

A2 VMs are the largest GPU cloud instances with NVIDIA A100 GPUs.

Recently, Google announced the GA release of A2 VMs based on the NVIDIA Ampere A100 Tensor Core GPUs in Compute Engine. The release enables you to run your NVIDIA CUDA-enabled machine learning (ML) and HPC scale-out and scale-up workloads more efficiently and at a lower cost. 

Google's A2 VMs provides 16 NVIDIA A100 GPUs in a single VM, which makes it the largest single-node GPU instance on the market today. The A2 VM also allows users to choose smaller GPU configurations (1, 2, 4 and 8 GPUs per VM),  providing the flexibility and choice a user needs to scale his/her workloads.

Source: Google

Google said that for ultra-large GPU clusters need, Google Cloud supports clusters of thousands of GPUs for distributed ML training and optimized NCCL libraries. The single VM shape offering with 16 A100 GPUs tied together with NVIDIA’s NVlink fabric is offerd by Google Cloud only. So, now you can start with one A100 GPU and go all the way up to 16 GPUs without having to configure multiple VMs for a single-node ML training. 

NVIDIA A100 GPU instances are now available in the us-central1, asia-southeast1 and europe-west4 regions. You can avail A2 Compute Engine VMs via on-demand, preemptible and committed usage discounts. A2 Compute Engine VMs are also fully supported on Google Kubernetes Engine (GKE), Cloud AI Platform, and other Google Cloud services.  You can get A100 GPUs for as little as $0.87 per hour per GPU on Google's preemptible A2 VMs.

You can get started training ML models, and serving inference workloads on NVIDIA A100 GPUs with Google's Deep Learning VM images in any of the available regions. These images also include all the software required like drivers, NVIDIA CUDA-X AI libraries, and popular AI frameworks.

Google said that it handles all software updates, compatibility, and performance optimizations, so users do not have to think about it.