(Original title: NVIDIA announced that it will build the world's fastest artificial intelligence supercomputer)

News from IT House, October 16 Nvidia announced yesterday that it will cooperate with the Italian University Research Center CINECA (one of the most important supercomputing centers in the world) to build the world's fastest AI supercomputer.

▲ Picture source NVIDIA

The new "Leonardo" system is based on Atos's BullSequana XH2000 supercomputer node, each node has four NVIDIA Tensor Core GPUs and an Intel CPU, and is expected to provide FP16 AI performance of hundreds of billions of triggers to support the integration of advanced AI and HPC Application use cases.

Leonardo will use Nvidia's Ampere architecture GPU and Nvidia's Mellanox HDR InfiniBand network, with nearly 14,000 Ampere GPUs, 1,800 HPC applications and AI applications, and will help scientists solve scientific challenges in many disciplines from materials science to high energy physics to climate change.

Leonardo is purchased by EuroHPC. EuroHPC is a cooperation between governments and the European Union to develop a world-class supercomputing ecosystem and exascale supercomputing in Europe, and is funded by the European Commission through the Italian University and Research Department.

IT Home learned that Nvidia will build four high-performance computers in Europe, including Supercomputing.

Source: IT House, translated by Google Translate