15,2 PetaFlop/s EuroHPC supercomputer in Ostrava at IT4Innovations to be delivered by HPE

14 Oct 2020 - Earlier this month, a contract to acquire the most powerful supercomputer in the Czech Republic, which will also rank among the most powerful in Europe, was signed. The contract was signed between the European High Performance Computing Joint Undertaking (EuroHPC JU), Hewlett Packard Enterprise, which will deliver the system, and IT4Innovations National Supercomputing Center VSB at the Technical University of Ostrava, where the new supercomputer will be located. The supercomputer will serve both academia and industry.

In the spring of 2019, IT4Innovations National Supercomputing Center achieved success with their project titled IT4Innovations Centre for European Science and Industry in a pan-European competition to host a European petascale supercomputing system, which will be built under the EuroHPC JU. A public procurement procedure to select the economic operator to supply the system was commenced, from which Hewlett Packard Enterprise emerged as the winner, becoming the supplier of the system with a theoretical peak performance of 15,2 PetaFlop/s. The signing of the contract between EuroHPC JU, VSB - Technical University of Ostrava, and Hewlett Packard Enterprise took place on October 2, 2020.

The cost of the procured system will total almost 15 million euro, 35 %, i.e. 5,13 million euro, of which will be paid by the EuroHPC JU. The remaining costs amounting to 9,73 million euro will be funded using the resources of the European Structural and Investment Funds through the Operational Programme Research, Development and Education (OP RDE). The installation and launching of this unique computing system are planned for the first quarter of 2021, and its service to science, industry, and society is planned to be provided until 2025. Its total operating costs between 2021 and 2025 are estimated at 14 million euro.

The supercomputer with working title EURO_IT4I will achieve a theoretical peak performance of 15,2 PFlop/s, thus becoming the most powerful supercomputer in the Czech Republic with an ambition to rank 10th in Europe and 50thin the world. It will thus exceed the performance of Salomon, the most powerful supercomputer at IT4Innovations to date, with a peak performance of 2 PFlop/s, reached many times. It is certainly worth noting that the accelerated part of the supercomputer, which will be made up of HPE Apollo 6500 servers with a total of 560 NVIDIA A100 Tensor Core GPUs that provide, among other things, a theoretical peak performance for artificial intelligence calculations of up to 175 PFlop/s. The supercomputer will be interconnected with NVIDIA Mellanox HDR 200Gb/s InfiniBand, delivering extremely low latency, and smart in-network computing acceleration engines. With NVIDIA GPU acceleration and smart InfiniBand offload engines, the new system will be perfectly prepared to handle very demanding, computationally intensive tasks in the fields of machine learning and artificial intelligence, which are delivering scientific breakthroughs across the globe.

The name of the new supercomputer will be selected as part of a public competition that everyone with permanent residency in the Czech Republic can participate in, and it will last until December 15, 2020.

Technical details:

The new supercomputer shall reach a theoretical peak performance of 15,2 PFlop/s, which corresponds to 15,2 quadrillion floating-point operations per second.

Technical data:

  • More than 100.000 CPU cores and 250 TB of RAM,
  • More than 3,8 million CUDA cores / 240.000 tensor cores of NVIDIA A100 Tensor Core GPU accelerators with a total of 22.4 TB of superfast HBM2 memory,
  • 15,2 PFlop/s total theoretical computing power,
  • Fast disk capacity with a capacity of 1 TB/s.

The supercomputer will consist of 6 main parts:

  • a universal part for standard numerical simulations, which will consist of approximately 720 computer servers with a theoretical peak performance of 3,8 PFlop/s,
  • an accelerated part with 70 servers and each of them being equipped with 8 GPU accelerators providing a performance of 11 PFlop/s for standard HPC simulations and up to 150 PFlop/s for artificial intelligence computations,
  • a part designated for large dataset processing that will provide a shared memory of as high as 24 TB, and a performance of 74 TFlop/s,
  • 36 servers with a performance of 131 TFlop/s will be dedicated for providing Cloud services,
  • a high-speed network to connect all parts as well as individual servers at a speed of up to 200 Gb/s,
  • data storages that will provide space for more than 1 PB of user data and will also include high-speed data storage with a speed of 1 TB/s for simulations as well as computations in the fields of advanced data analysis and artificial intelligence.

Ad Emmen / IT4Innovations