NVIDIA must be familiar to everyone. Many OEMs, graphics card manufacturers, system manufacturers, and consumer electronics companies around the world have chosen NVIDIA’s processors as the core components of their entertainment and business solutions. Nvidia said it would launch a new chip, the A800, in China, in line with recent U.S. export control regulations. The spokesman said the A800 chip will be put into production in the third quarter as a replacement for the A100 chip, which has been restricted from exporting to China by the U.S. Department of Commerce.
The A100 uses the NVIDIA Ampere architecture, the engine behind the NVIDIA data center platform. The A100 offers up to 20x performance improvement over the previous generation and can be partitioned into seven GPU instances to dynamically adjust to changing needs. The A100 is available in 40GB and 80GB memory versions, the A100 80GB doubles the GPU memory and provides ultra-fast memory bandwidth (over 2 terabytes per second [TB/s]) for processing very large models and data set.
A100
The A100 is part of an overall NVIDIA data center solution, which is an overlay of hardware, networking, software, libraries, and optimized AI models and applications in NGC. With Tensor Floating Point (TF32) precision, the A100 Tensor Core delivers 20x better performance than NVIDIA Volta without code changes; with automatic mixed precision and FP16, the performance can be further improved by 2x.
