Contract for Leonardo pre-exascale 248 Petaflop/s EuroHPC supercomputer signed

Eugenio Di Sciascio, President of CINECA.
15 Oct 2020 Bologna - The European supercomputer Leonardo, which will be managed by CINECA and will be installed in 2021 at the Tecnopolo in Bologna, will be based on Atos BullSequana XH2000 technology, and will be equipped with approximately 14.000 GPU accelerated processors from NVIDIA. It will have an aggregate HPL LINPACK performance of 248 Petaflop/s and a storage capacity of over 100 petabytes. This is about 10 times the computing power of CINECA's Marconi100 system, currently in ninth place in the world ranking of TOP500 supercomputing systems. Today, this would be the second fastest supercomputer in the world, after the Fugaku. The system is committed to appear in the TOP500 in November 2021.

The Leonardo contract was announced on October 15, 2020 by the Minister of University and Research Gaetano Manfredi; Roberto Viola, Director General of DG Connect at the EU Commission; the President of CINECA Eugenio Di Sciascio; the DG of CINECA David Vannozzi in a press conference at which also participated Thomas Skordas, director of "Digital Excellence and Science Infrastructure" DG CONNECT European Commission; Sanzio Bassini, HPC CINECA Director; Anders Dam Jensen, Executive Director EuroHPC Joint Undertaking; Antonio Zoccoli, INFN president; Stefano Ruffo, SISSA president; Giuseppe Di Franco, CEO Atos Italia; and Marc Hamilton, vice president of solutions architecture and engineering at NVIDIA.

As part of the strategy implemented by the joint EuroHPC initiative to finance three supercomputing systems of the pre-exascale class, this is the first announcement relating to the allocation of technology suppliers. Italy's national strategy - the other two systems will be hosted by Finland and Spain - is centred on a mandate to CINECA to act as hosting entity according to an agreement between the Ministry of University and Research, CINECA, INFN and SISSA and in synergy with the Emilia-Romagna Region for the data centre. The Hosting Entity Consortium sees the participation of Slovenia, Slovakia, Austria and Hungary, and soon also Greece.

The LUMI contract for the pre-exascale system in Finland will be announced later this week.

Italy has a long tradition of success in the field of supercomputing, thanks to the foresight of MIUR and the operational capacity and expertise of CINECA, to guarantee the national and international scientific community an adequate and competitive supercomputing infrastructure.

This European and Italian investment will make it possible to provide European researchers with a supercomputer comparable to those installed by the United States, China and Japan, which will rank at the top 5 of the TOP500 list of the most powerful supercomputers in the world, the European Commission expects.

High-performance computing is a strategic tool for the challenges arising from competitive science, industry and public decision-making, in fields ranging from fundamental physics, climate research, biology, life and health sciences, up to supporting public decision-making in the event of natural risk events. To solve these challenges, it is increasingly necessary to rely on the synergy of advanced algorithms, data and hardware based on the next generation of exascale supercomputers.

In recent years, the United States, China and Japan have taken the lead in the HPC area, with supercomputing considered as enabling the economic success of countries. Through the initiative of the EuroHPC Joint Undertaking, Europe has decided to invest in European exascale advanced computing infrastructures to meet the growing computing and data needs of European scientists and industry. The aim is to pool European resources to develop high-end exascale supercomputers based on competitive European technology. Atos is a European industry in the field of HPC technology producers, and CINECA's goal is to establish an international reference point for technological supervision and the development of European technology for supercomputing, a strategic asset for independence and the competitiveness of the entire European as well as the national level.

Modern scientific computing requires HPC simulation, data analysis, artificial intelligence, and machine learning and visualization. The NVIDIA Compute Platform accelerates all of these workloads by providing extremely high throughput and low power consumption, making it ideal for scientific computing. Research examples using this approach include work in areas such as drug discovery, space exploration and research, weather modelling.

Thanks to the Leonardo system, researchers will be able to identify strategies for mitigation and response to natural events, such as earthquakes, tsunamis, volcanic events, hydrogeological assets, to combat pandemic situations and pathogenic epidemics. In this regard, the road towards Italy's leadership role has already been started in the context of the ongoing emergency, in the reference projects of the European anti-COVID-19 action, such as the Escalate4CoV project for the reuse of active and synthetic ingredients that are already known in the preparation of new drugs for contrast therapy against COVID-19, and the Orchestra project, for the management and monitoring of contagion clusters in the fight against the spread of the pandemic. In this context, among the other priority issues that have emerged so far are quality of the environment, cybersecurity and Artificial Intelligence.

Eugenio Di Sciascio, CINECA President, stated at the press conference: "Italy participates, as a founding country, in the EuroHPC Joint Undertaking, whose mandate is the implementation in Europe of a world-class supercomputing infrastructure and a competitive innovation ecosystem in technologies, applications and, above all, in skills and competences in high performance computing. Our country, in fact, has a long tradition of success in this area thanks to the foresight of the Ministry of University and Research, the persistent operational capacity and expertise of CINECA, to guarantee the national and international scientific community a supercomputing infrastructure that is adequate and competitive. CINECA has been a reference point for the national and international scientific community in high-performance scientific computing technologies for over 50 years. Thanks to state-of-the-art IT resources and excellent skills, it represents Italy at an international level and has always provided state-of-the-art solutions in computational science to public and industrial research."

David Vannozzi, CINECA CEO, stated: "It is with a hint of pride that we announced today the suppliers of the Leonardo supercomputer. Thus, the first stone of an ambitious building was laid that will enable Italy and Europe to compete with China, the United States and Japan in the field of supercomputing. This is an extraordinary commitment from the European Commission and the 32 participating States. It should be remembered that supercomputing is today a decisive booster in economic competition, and beyond. Supercomputing is also the gateway to artificial intelligence and with the supercomputer Leonardo, belonging to the exascale class, elements of knowledge are put at the disposal of the political decision-making in real time. What is already happening in the field of medical-scientific research to combat COVID-19, with the isolation of an 8n molecule capable of impacting the coronavirus spike protein, can also be repeated in other fields. I am thinking of the presence of one of the most important weather centres here in Bologna, and of the possibility of enhancing, thanks to supercomputing, weather forecasts with increasingly detailed maps referring to very small portions of the territory. The areas of application of supercomputing are unlimited. Today we have laid the foundations to enter the fourth industrial revolution and "re-read" reality, to modify it, through supercomputing."

Roberto Viola, Director General of the Directorate General for Communications Networks, Content and Technology (DG Connect) of the European Commission, stated: "I am very pleased with the announcement of the Leonardo pre-exascale supercomputer which will be developed and installed at CINECA by a European supplier in the coming months. This supercomputer will be among the top five in the world. It will combine the best of Artificial Intelligence and HPC technologies. It will help further develop the capabilities of Italy and Europe in a number of key application areas, such as responding to climate change, increasing resilience to natural disasters and pandemics and promoting the research and innovation capacities of our sector."

Anders Dam Jensen, EuroHPC JU Executive Director, stated: "Today is a concrete and promising milestone for European supercomputing. Leonardo is the first EuroHPC pre-exsascale procurement contract to be announced. It will support Europe to lead the way in the technological race to exascale, benefiting European researchers, industry and the public sector in innovation, bringing unimaginable benefits to citizens in fields such as medicine, energy and agriculture, to name but a few."

Giuseppe Di Franco, Group Executive Vice President, CEO of Atos in Italy, stated: "As the European leader in high performance computing, Atos is proud to be part of this project, for Europe and for Italy. Supercomputers, with their ability to handle millions of billions of operations per second, open up endless possibilities in the fields of scientific, industrial and medical research and are the basis for ensuring that all citizens can benefit from the benefits of digital transformation."

"The EuroHPC technology roadmap for exascale in Europe is opening the door to rapid growth and innovation in high performance computing and AI", stated Marc Hamilton, vice president of solutions architecture and engineering at NVIDIA. "We are working with CINECA and Atos to accelerate scientific discovery across a wide range of application domains, providing a platform to usher in the era of exascale computing."

CINECA's Leonardo supercomputer will be based on the Atos BullSequana XH2000 supercomputer. With an aggregate HPL performance of 248 Petaflop/s, it will be an order of magnitude more powerful than the current best performing supercomputer in Europe and with a storage capacity of over 100 petabytes. Designed, developed, manufactured and tested in Europe, the Atos supercomputer will provide Italian and EuroHPC researchers with a modern infrastructure that offers maximum performance for computing, data processing, High Performance Data Analytics (HPDA), Artificial Intelligence (AI) and the development of Machine Learning models, enabling them to overcome the limitations of traditional simulations and explore new applications, accelerating transformation and innovation.

Leonardo will be built from Atos' BullSequana XH2000 supercomputer nodes, each with four NVIDIA Tensor Core GPUs and a single Intel CPU. It will also use NVIDIA Mellanox HDR 200Gb/s InfiniBand connectivity, with smart in-network computing acceleration engines that enable extremely low latency and high data throughput to provide the highest AI and HPC application performance and scalability.

Leonardo will feature nearly 14.000 NVIDIA Ampere architecture-based GPUs. It will deliver 10 exaflops of FP16 AI performance. NVIDIA Ampere architecture GPUs can accelerate over 1800 applications such as Quantum Espresso for material science, SPECFEM3D for geoscience and MILC for quantum physics by up to 70x, making previous big challenge simulations almost real-time tasks.

Third Generation Intel Xeon Scalable processors (Ice Lake) are optimized to perform computationally intensive workloads in high-performance computing systems like Leonardo. The follow-on processor to Intel's Ice Lake server processors is Sapphire Rapids, which will enable exascale computing with advanced built-in AI acceleration capabilities.

Technical features of the Leonardo include the following:

  • More than 150 BullSequana XH2000 liquid-cooled rack
  • 248 LINPACK performance Petaflop/ss (R-max)
  • 1536 CPU nodes with Intel Sapphire processors
  • 3456 servers equipped with next generation NVIDIA Ampere GPUs
  • 5 Petabyte of memory
  • 100 Petabyte of storage
Ad Emmen