While gamers continue to struggle to get their hands on Nvidia’s consumer-grade Ampere GPUs, an inter-university consortium could dive into a vault of GPUs ala Scrooge McDuck.

CINECA, a supercomputer centre, will be making use of a new system called Leonardo that will be housed in Italy. The computer is expected to deliver 10 exaflops of half-precision floating-point (FP16) artificial intelligence (AI) performance.

How? Well for one Leonardo will contain 14 000 “Nvidia Ampere architecture-based GPUs”. The supercomputer will also feature Nvidia’s Mellanox HDR 200Gb/s InfiniBand networking.

Drilling a bit deeper, Leonardo will be built from Atos BullSequana XH2000 supercomputer nodes. Each node contains four Nvidia Tensor Core GPUs and a single Intel CPU along with the aforementioned Mellanox networking.

So why does one need all of this compute power?

These days scientific computing is incredibly demanding in terms of performance. Simulations, data analytics, AI and machine learning require a lot of compute power.

Leonardo will help scientists with drug discovery, space exploration, high-energy physics, climate change and so much more.

But Leonardo is not the only supercomputer worth mentioning. There were a total of four new supercomputers unveiled this week with another four coming later this year.

“All four supercomputers announced use Nvidia Ampere architecture GPUs and Nvidia Mellanox HDR InfiniBand networks to tap an ecosystem of hundreds of HPC [high performance computing] and AI applications. Atos, an Nvidia systems partner headquartered in France, will build three of the four systems; Hewlett Packard Enterprise will construct the fourth,” Nvidia wrote in a blog post.

Aside from Leonardo there is MeluXina in Luxembourg which will connect 800 Nvidia A100 GPUs on HDR 200Gb/s InfiniBand links.

The Vega supercomputer in Slovenia will include 240 A100 GPUs.

“The IT4Innovations National Supercomputing Center will host what’s expected to become the most powerful supercomputer in the Czech Republic. It will use 560 NVIDIA A100 GPUs to deliver nearly 350 petaflops of AI performance,” wrote Nvidia’s head of Solutions Architecture and Engineering team, Marc Hamilton.

All of these supercomputers will form a network driven by EuroHPC with a view to creating “an engine to power Europe’s data economy”.

With all of that computing power, here’s hoping we can tackle some of the world’s most pressing problems.

[Image – Nvidia]