Best-performing supercomputer for AI research in CZ

We have joined forces with the Faculty of Electrical Engineering CTU and together we have established the Research Center for Informatics (RCI). The aim of the project is to build the most powerful computer cluster for research into artificial intelligence in the Czech Republic worth CZK 41.6 million. This unique piece of equipment with supercomputer performance is located in the basement of the historical building of the Faculty of Electrical Engineering CTU on Karlovo Square, which brings many challenges.

The Research Center for Informatics RCI is a leading scientific workplace in the Czech Republic in the field of computer science and artificial intelligence. The centre’s aim is to further develop competitive and high-quality research at international level, boost collaboration between basic and applied research, welcome leading scientists from abroad and connect senior scientists and students. RCI is funded from the Operational Program Research, Development and Education, call Excellent Research, with a total budget of CZK 580 million. The grant provided by the Ministry of Education, Youth and Sports made it possible to create a cluster for research in the field of artificial intelligence with the best performance in the Czech Republic. The equipment will be further upgraded in 2021.

The centre is led by Prof. Dr. Michal Pěchouček, MSc., head of the Department of Computer Science, Faculty of Electrical Engineering, CTU in Prague, who says:

“Today, we have a unique opportunity to work on joint research across the university and take the development of IT to international level. Experts at the centre create value in machine learning, artificial intelligence, theoretical computer science, bioinformatics, high-performance computers, big data, cybersecurity, computer graphics and more. The newly built cluster also opens up new possibilities for us that were previously out of our reach.”

Also, research teams at the Faculty of Information Technology welcome the acquisition of a high-performance cluster which CTU has called for in the last few years.

“Research and experimental development of high-performance and scalable algorithms is conducted in three phases. First on personal computers for small data instances, after which the correctness and scalability of algorithms are checked on a sufficiently big cluster. Afterwards, the efficiency and scalability are evaluated on the world’s biggest supercomputers. The top architecture of the newly installed RCI cluster will enable researchers to efficiently perform phase II of the development of highly scalable algorithms,” explains prof. Ing. Pavel Tvrdík, CSc., head of the Department of Computer Systems, the Faculty of Information Technology, CTU in Prague.

The equipment has unique technical specifications.

“NVIDIA V100 Tensor Core GPU is the best performing accelerator for high-performance computing and artificial intelligence. The total installed output of over 6 PetaFLOPS makes the CTU installation currently the most powerful supercomputer for AI applications in the Czech Republic,” says Rob Evans, EMEA director for high-performance computing and artificial intelligence at NVIDIA.

The most powerful supercomputer for AI applications in the Czech Republic.

Putting this high-performing supercomputer in operation was a challenge.

“The entire installation took six weeks, including the delivery of all hardware components, construction work, air conditioning installation and software installation,” explains Petr Plodík at M Computers, who was in charge of implementing the cluster. “Building such a cluster in the centre of Prague is quite unique. We had to create new wiring for power and cooling, and we also faced the limits of the local transformer station,” he adds.

The cluster consists of 20 Gigabyte CPU nodes with Intel Xeon Gold processors (480 processor cores in total), 12 NVIDIA GPU Supermicro computing nodes, one Lenovo ThinkSystem SR950 node with a high number of CPU cores and shared memory (192 processor cores, 1.5 TB memory), high-speed Infiniband EDR interconnecting networks (100Gb/s) from Mellanox, super-fast NVMe SSDs from Western Digital and a DELL EMC Isilon shared scalable disk array. Each GPU node is equipped with 4 NVIDIA V100 Tensor Core GPU accelerators. In total, the cluster has 48 accelerators with a total of 245,760 CUDA cores, 30,720 Tensor Cores and total computing power of over 6 PetaFLOPS in artificial intelligence operations.

The cluster has a total computing power of over PetaFLOPS in AI operations.

With this high-performance cluster, researchers at RCI will be able to conduct basic research into deep learning methods that are key to artificial intelligence, such as applications for robotics, bioinformatics, development of applications for computer security and control of autonomous cars. Thanks to the RCI project, the Czech Republic will now be able to compete in these areas with universities and centres of excellence around the world.