Nvidia strengthens its offer in the professional segment with the announcement of a variant of its A100 GPU. It offers twice as much storage space while taking over HBM3E.
The manufacturer decides to move up a gear with a better armed A100-amp GPU. This version benefits from twice as much memory with 80 GB HBM2 instead of 40 GB HBM2. This change allows for a higher speed of 3.2 Gbps instead of 2.4 Gbps, which is enough to bring the tape up to 2 TBps. For the rest, the presence of 6,912 CUDA FP32 cores and a GPU frequency at 1.41 GHz does not change anything.
A100 Ampere, a graphics processor equipped for computationally intensive calculations
This GPU is not intended for our PCs. It is designed for the world of AI intensive computers. In the next months, Nvidia intends to offer both versions. This new A100 is not intended to replace the current version. It will take place in HGX and DGX configurations equipped with 4 or 8 GPUs.
source: Nvidia