Nvidia chips and equipment to power world’s fastest AI supercomputer

Nvidia chips and equipment to power world’s fastest AI supercomputer.
YouTube | Cineca

Nvidia announced its products would be used to power “Leonardo,” the world’s fastest artificial intelligence (AI) supercomputer. French information technology firm, Atos, is building the system, and Cineca, an Italian inter-university group, will operate and maintain it.

Once online, Leonardo will offer 10 exaflops of half-precision floating-point (FP16) AI processing power.

How Leonardo Came Into Existence

In 2016, European Union leaders decided to fund a new initiative to increase the region’s next-generation computational resources. Two years later, the European High-Performance Computing Joint Undertaking (EuroHPC), a public and private sector initiative, launched to tackle the mission.

Manage your supply chain from home with Sourcengine

Backed with a €1.4 billion budget, the organization decided to create a cluster of interconnected supercomputers throughout the EU. Leonardo, which will be maintained in Cineca’s Bologna, Italy campus, is one part of the EuroHPC cluster.

Post activation, the Italian machine will be linked with Luxenberg’s Meluxina, Slovenia’s Vega, and planned high-performance systems in Finland, Portugal, Bulgaria, and Spain.

Nvidia’s Contribution to the Future of Supercomputing

Atos will utilize Nvidia’s Ampere graphics processor units (GPUs) and Mellanox HDR InfiniBand networking switches to construct Leonardo.

14,000 cutting-edge data center components will supercharge the system’s AI-enabled processing capability. It will also use InfiniBand switches to link Leonardo to its siblings because of their 200Gbps throughput and low latency. The firm’s BullSequana XH2000 supercomputer nodes, each housing four Nvidia Tensor Core GPUs, will act as the machine’s foundation.

Cineca intends to use Leonardo on a host of data-intensive research projects.

The organization will perform advanced genomic analysis tasks to find new drugs to treat the coronavirus and other illnesses. The group will also collate information from a range of sources to conduct high-level astrophysics inquiries. In addition, researchers will use the incredibly powerful AI platform to forecast extreme weather events.

In a press release, Nvidia explained its GPUs would enable Leonardo to run quantum physics applications 70 times faster than prior generation systems.

As far as chip sector evolutions go, Nvidia’s is unique in its speed and scale. A few years ago, the firm developed components that made video games look breathtaking. Soon, its GPUs and networking hardware will drive innovation in several scientific disciplines.


Please enter your comment!
Please enter your name here