Marconi100 offers a processing power of almost 32 theoretical peak petaFlops, i.e. up to 32 million billion calculations per second.
he Marconi100 project consolidates the multi-year collaboration between IBM and Cineca , the Interuniversity Consortium for High Performance Computing formed by the Ministry for University and Scientific Research, 69 Italian universities and 11 national public institutions.
The Marconi100, the 9th most powerful supercomputer globally, and 2nd in Europe, as attested by the latest edition of the Top500 ranking, was built with the same IBM Power9 technology and NvidiaV100 Tensor Core GPUs as Summit, the most advanced for the era of artificial intelligence , now under the US Department of Energy.
According to the measurements made by Cineca, Marconi100 offers a processing power of almost 32 theoretical peak petaFlops, i.e. up to 32 million billion calculations per second.
Cineca thus reconfirms its mission of support to the national and European community in the area of supercomputing with a Tier0 class system, the highest classification on a continental level. Marconi100 will in fact be able to support European researchers, through the Prace initiative, and Italian researchers through the Iscra (Italian Supercomputing Resource Allocation) initiative, to face the socio-economic challenges of our time such as climate change, renewable energy, sustainable economy, precision medicine.
Furthermore, by drastically reducing digital simulation times, supercomputing is able to accelerate knowledge and decision support and can be of great help in dealing with risk management situations, such as measuring environmental impact, or crisis situations such as earthquakes or other extreme events such as the current pandemic.
In particular, one of the first projects for Marconi100 is the current use in an anti-Covid19 key to identify therapies capable of reducing the pathogenic aggressiveness of the coronavirus: for example, by digitally simulating the behavior of proteins that allow the virus to replicate better. identify molecules and pharmaceutical compounds capable of inhibiting it.
Cineca is already a partner of many active projects both nationally and internationally and participates in the Exscalate4CoV project, funded with an emergency procedure by the European Commission.
How does AI help optimize business operations?
- By checking this box, I confirm that I want to receive marketing communications and information on products, services and events from OpenText. I am aware that I can cancel my membership at any time. For further details on how OpenText shares, protects, stores and transmits personal data, see: OpenText Privacy Policy. Our cookie policies can be found here.
“The Marconi100 project – declared Enrico Cereda, president and CEO of IBM Italy – is part of a long, fruitful cooperation with the Interuniversity Consortium, also involving our centers of excellence in Europe. Equipped with the same technology as two of the top three supercomputers at the top of the world rankings, Marconi100 will follow in its footsteps by offering the research community the computing power necessary to accelerate the discovery of a cure for Covid-19. As the activity of the Summit system shows in particular, without the availability of these architectures it would take much longer times, incompatible with the urgent need to find a solution. We are proud to be alongside Cineca, and the entire country, in this and other important scientific challenges “.
“With the installation of the Marconi100 partition – said the acting president of Cineca Eugenio Di Sciascio – the Marconi project for the persistent evolution of the Cineca calculation system started in 2016 is completed which, today, allows us to have a aggregate computing power of approximately 50 peak petaflops. The Marconi100 partition, against the technological maturity derived from the experience acquired by IBM with the Summit and Sierra projects, will allow us to guarantee the national and European scientific community access to a super-computing system of great production stability and high quality. of the service “.
“This confirms the centrality of the Consortium in the development of research in the academic and industrial fields – underlined the director general of Cineca, David Vannozzi – as happens with Exscalate4CoV, a project led by the pharmaceutical company Dompé, engaged in the research and evaluation of proteins. for the fight against coronavirus “.
Cineca is a partner of over 20 international projects for the development of the European HPC infrastructure and its technological and application ecosystem: in the field of plasma physics with the Eurofusion project, whose goal is to bring about significant progress in the production of fusion energy, in the science of matter and materials with the MAX (Material Science at Exascale) project, industrial engineering with the Exellerat Engineering project and solid earth physics with the Cheese Solid Earth project. All projects that will benefit from the high processing capacity provided by Marconi100.
The design and integration of the Marconi100 system are the result of IBM’s experience in the construction of the Summit and Sierra. Similar to both, the Marconi100 architecture uses IBM Power9 processors and Nvidia Tensor Core V100 GPUs as well as the high-performance Nnvidia Mellanox InfiniBand interconnect in a high-performance, highly scalable DragonFly + topology. Marconi100 also benefits from eight petabytes of IBM Elastic Storage Server with IBM Spectrum Scale to support a high performance read / write load and offer a balanced compute and input / output architecture.
The use of an accelerated computing architecture is expected to improve the productivity of many applications, with a significant increase in performance at both the single server node and aggregate cluster levels connected across the high-performance Infiniband network.
In terms of energy sustainability, the increase in processing power will not lead to an increase in consumption: the new system will be able to count on greater energy efficiency, with the same processing capacity, by a factor equal to about three times the pre-existing system.
