NIVIDA Ampere Architecture GPU Range Explained

NVIDIA pioneered accelerated computing with GPU processors. Their newly expanded range of Ampere Architecture GPU accelerators provide a range of PCIe based GPU engines to accelerate a wide variety of data intensive workloads in the data centre.

More data workloads are moving to GPUs because of speed as well as energy efficiency and space efficiency in the data centre. By moving these data intensive workloads to the GPU, a smaller number of servers is required, less power is consumed, and processing jobs are completed quicker! A win-win-win!

NVIDIA’s data centre platform is the world’s leading GPU acceleration tool, utilising NVIDIA GPU Cloud (NGC) to containerise workloads, enabling scale and easy management of data centre infrastructure. Two of the new Ampere Architecture GPUs feature NVIDIA’s Multi-Instance GPU (MIG), a unique feature that allows a GPU to be split into separate logical units, or combined across a number of GPUs to make one large logical unit. Similar to virtualisation, but running at a hardware level with each MIG slice allocated dedicated memory and processing capabilities. MIG makes the new Ampere Architecture GPUs agile, enabling them to be configured for small workloads, combined for larger requirements, and shared across a team.

The new Ampere Architecture Range

As a general guide, the new range of Ampere Architecture GPUs are designed for specific workloads, and combinations of multiple GPUs.

A100 – highest performance computing GPU, ideal for Deep Learning training and Data Analytics, as well as Inference, High Performance Computing (HPC), accelerating enterprise workloads and edge applications. The A100 also supports MIG.
A30 – the mainstream computing GPU, the A30 is a workhorse for High Performance Computing (HPC), and artificial intelligence and deep learning work. The A30 can also be deployed to accelerate enterprise applications and edge AI workloads. The A30 also supports MIG.
A40 – the high performance graphics GPU, the A40 is ideal for render farms, graphics workloads, cloud gaming as well as enterprise and edge applications.
A10 – the mainstream graphics GPU, the A10 delivers fantastic acceleration and efficiency in graphics work, cloud gaming, and enterprise and edge applications.
A16 – optimised for Virtual Desktop Instances (VDI), the A16 is designed to support graphics workloads for a large number of users, making GPU acceleration a cost effecting solution for a team.