Nvidia Unveils A100 GPUs Based On Ampere Architecture

Nvidia on Thursday announced a bevy of new products and company updates via a virtual GTC Technology Conference keynote address from founder and CEO Jensen Huang. Key updates include the launch of the A100, Nvidia's 8th generation GPU design and its first based on Ampere architecture.

Nvidia said the A100 represents the biggest generational leap ever for one of its GPUs. Designed for data centers, the multi-instance GPU is optimized for HPC and inference, delivering 20x speed improvements over Volta, with more than 54 billion transistors and third-generation Tensor Cores.

Also: Ampere launches Altra, takes aim at AMD, Intel in cloud data centers 

With the A100, Nvidia said machines will be capable of processing massive amounts of data very quickly, and the servers will be more flexible.

"It's going to unify that infrastructure into something much more flexible and increase its utility, which makes it a lot easier to predict how much capacity you need," Huang said Wednesday during a press briefing. 

The A100 is in full production and already shipping to customers worldwide. Cloud computing giants including Amazon Web Services, Alibaba, Baidu, Cisco, Dell Technologies, Google, HPE, Microsoft and Oracle are currently incorporating the GPU in their data center architectures, Huang said.

The Jetson Xavier NX developer kit is designed for drones, delivery robots, and anything with a small form factor, the company said. The system was first announced in December and is now generally available. 

The company is also shipping a third generation of its Nvidia DGX AI system based on the A100. Huang said the Nvidia DGX A100 is the world's first 5-petaflops server that can be divided into as many as 56 applications running independently. Nvidia said initial recipients of the system include the US Department of Energy's Argonne National Laboratory, which will use the cluster's AI and computing technology for COVID-19 research.

Also: Nvidia and Databricks announce GPU acceleration for Spark 3.0 

Also on the healthcare side, Huang said Nvidia has been working with researchers and scientists to offer its GPUs and AI computing technology as part of the pandemic response.  Among those mentioned are the Oak Ridge National Laboratory and the Scripps Research Institute. The organizations are using advanced computing to screen a billion potential drug combinations each day, Nvidia said. Nvidia also announced updates to its Clara healthcare platform aimed at taking on COVID-19. 

"Researchers and scientists applying NVIDIA accelerated computing to save lives is the perfect example of our company's purpose — we build computers to solve problems normal computers cannot," Huang said. 

On the autonomous machines side, Nvidia announced that German automaker BMW is using Nvidia's robotics platform to redefine factory logistics. BMW built an end to end system on the Isaac platform, training five AI robots using ray tracing GPUs to render machine parts. 

Huang also unveiled Merlin, an SDK to build AI-based recommendation systems. Merlin reduces the time it takes to create a recommender system from a 100-terabyte dataset to 20 minutes from four days, Huang said. The company also detailed the latest developments on Jarvis, the company's conversational AI SDK first announced in October.