Cineca Uses Nvidia Accelerators To Build World’s Fastest AI Supercomputer

Cineca Uses Nvidia Accelerators To Build World’s Fastest AI Supercomputer

Italian inter-university consortium Cineca, which has a large supercomputer center, plans to create the world’s fastest supercomputer for artificial intelligence (AI).

The new Leonardo system is expected to deliver 10 exaflops (1018 floating point operations per second) on FP16 data operations in artificial intelligence applications. The system will have nearly 14,000 Nvidia GPUs based on Ampere architecture. More specifically, it will be built from Atos BullSequana XH2000 nodes with four Nvidia GPUs and one Intel processor each. Nvidia Mellanox InfiniBand HDR solutions with 200 Gbit / s bandwidth will be used for internal connections.

The system will be supplied by EuroHPC, a public-private high-performance computing partnership between national governments and the European Union to create a world-class supercomputer ecosystem and develop supercomputing in Europe. The project is funded by the European Commission through the Italian Ministry of Education, Universities and Research.

Modern scientific computing requires high-performance modeling, data analysis, artificial intelligence, machine learning and visualization, according to Nvidia. The Nvidia Computing Platform accelerates all of these workloads with extremely high bandwidth and low power consumption, making it ideal for scientific computing.

The supercomputer Leonardo will help solve scientific problems in many disciplines, from materials science to high energy physics and climate change. Scientists and researchers will immediately be able to use it with great efficiency, as it will work with the same CUDA software as the existing system installed at Cineca, which is the fastest research supercomputer in Europe used by universities.

Cineca Uses Nvidia Accelerators To Build World’s Fastest AI Supercomputer

Leave a Comment