Petra Škaberna
The Faculty of Electrical Engineering at University of Zagreb (FER) is proud to announce that it has provided a significant contribution to European sovereignty, with valuable business impacts. European High Performance Computing Joint Undertaking (EuroHPC JU) finalized the 273-M€ procurement contract for JUPITER, the first EuroHPC exascale supercomputer granted to the consortium of Eviden and ParTec. JUPITER will be hosted at Forschungszentrum Jülich in Germany, where the brain of the system will be based on the brand new general-purpose Rhea processors by SiPearl that have been designed in Europe with FER being one of the key partners of SiPearl in design and verification of those processors.
"FER is proud to be at the forefront of what is among the most advanced global supercomputer processor technology research and to partner with leading European industry and research organizations SiPearl, Eviden, Barcelona Supercomputing Center and others to create the first European general purpose processor for supercomputers. Partnering with SiPearl allowed us to have a unique opportunity to be able to work with great industry experts on the design and verification of what will be the family of globally one of the most advanced and powerful processors.
Also, being a member of European Processor Initiative project funded by EuroHPC JU, we team with the brightest minds, share ideas and solutions and make Faculty of Electrical Engineering at University of Zagreb recognized on the global research map" , said Mario Kovač, professor and Director of FER HPC research center .
What are exascale supercomputers
Supercomputers solve problems that are significantly beyond the capacity of a single office computer/laptop, since they can process significantly larger amounts of data and make calculations in seconds that would otherwise take months or even decades. Supercomputers are built out of the most advanced components and use large and complex building blocks.
One way of describing the power of computers is by comparing how many basic operations, such as e.g. additions, they can compute in one second. Thus, the metric is given in so called FLOPS (floating point operations per second).
An "Exascale supercomputer" can perform more than billion billion or 1,000,000,000,000,000,000 (10 18 ) operations (FLOPS) per second, or in short more than one exaFLOPS. To put this into perspective: it would take one million typical laptops to do the same number of computations in the same amount of time.