NVIDIA Ampere GA100 GPU to feature 54 billion transistors.
NVIDIA Tesla A100
This is the world’s largest GPU and NVIDIA’s first manufactured in the 7nm fabrication process. NVIDIA GA100 features 54 billion transistors, which more than twice of what Volta GV100 had to offer (21B).
NVIDIA claims that the GPU will be 20x more powerful than Volta GPU in AI-training (single-precision operations) and AI interference (8-bit integer operations). In terms of high-performance computing (double precision operations) NVIDIA says that the GPU will be 2.5 faster.
“With this new precision, A100 offers 20 times more compute for single-precision AI, and because developers can continue to use the inputs as single-precision and get outputs back as single-precision, they do not need to do anything differently. They benefit from this acceleration automatically out of the box,”
- NVIDIA Kharya, director of product management for accelerated computing at Nvidia
NVIDIA Ampere features third-generation Tensor cores called TF32 designed for AI. New Tensor Cores now support FP64 numbers, which is a key upgrade for HPC applications. It means that the GPU will have much higher double-precision performance.
NVIDIA DGX A100
The DGX A100 system features eight Tesla A100 accelerations, offering up to 5 petaflops of performance. In NVIDIA’s own numbers, each of the Teslas offers 20x more peak performance than Volta-based DGX systems.
While the former generation was designed only with AI-operations in mind, DGX A100 is also scale-up applications (data analytics) and scale-out applications (inference).
NVIDIA is happy to demonstrate that a single rack of DGX1 system replaced 25 data center racks with CPUs, but only requiring 1/20 of the power and 1/10 of the investment.
NVIDIA also confirmed that the DGX A100 systems are already in operation in the US Department of Energy (Aragone National laboratory), where they are used to fight Covid-19.
Each of the DGX-A100 systems will cost 200,000 USD. The systems are already shipping.
Source: EETimes (removed)
