The Lenovo ThinkSystem servers will become a reference platform of the powerful supercomputer
Customers: SURF Amsterdam; Science and education Contractors: Lenovo Data Center Group Product: Lenovo ThinkSystemSecond product: Nvidia HGX Third product: Drafts of supercomputer platforms Project date: 2021/02
|
2021: Start of creation of the supercomputer based on the Lenovo ThinkSystem servers
On February 8, 2021 the Lenovo Data Center Group company (Lenovo DCG) reported that it provided high-performance supercomputer infrastructure of the Netherlands association of educational and research institutions on development of ICT of SURF. The project on development of the supercomputer estimated at 20 million euros will be started at the beginning of 2021.
According to the company, infrastructure will cover more than 100 educational and research institutions working in the most different industries and requiring high-precision calculations including in meteorology, astrophysics, medicine, materials science and geology.
Recently in the scientific world the shortage of computing powers, devices for storage and data processing is especially sharply felt. The project of the supercomputer is aimed first of all at the solution of research tasks. We selected Lenovo for quality, performance and flexibility of solutions and also for commitment to the principles of sustainable development. Walter Liyun, the manager on scientific and research of SURF association told |
Supercomputer infrastructure of Lenovo will include servers Lenovo ThinkSystem based on processors AMD EPYC the second generation and AMD EPYC of the next generation which will arrive in sale under the name Milan in the nearest future. All of them are equipped with technology liquid coolings Lenovo Neptune and also the solution Lenovo Distributed Storage Solution (DSS-G; 12.4 PiB) and NVIDIA HGX the A100 (4-GPU) platform — for work with artificial intelligence and machine learning. Smart switches NVIDIA Mellanox HDR InfiniBand (200 Gbps) which support calculations in network (in-network) will provide low delays and proper network transmission capacity. Thanks to peak capacity in 13 petaFLOPS and also the power, by 10 times exceeding an indicator of the previous system, infrastructure of Lenovo will become the tool in hands of researchers for many years.
The supercomputer should be not only highly productive, but also energy-efficient. The efficiency of technology of liquid cooling of Lenovo Neptune corresponds to 90% that will allow to reduce electricity consumption at peak values of system performance.
For us honor to take part in such project which not only will enrich research history in the Netherlands, but also will provide more energy efficient system — thanks to our last technology of liquid cooling. Tina Borbyer, the CEO of division of Lenovo DCG on work in Benelux countries and Northern Europe told |
At the heart of the graphic processors A100 with tensor cores involved in the project the architecture of NVIDIA Ampere lies. It accelerates training and work of ready algorithms of artificial intelligence in real life (inference) and also allows to carry out tasks in the world of AI and HPC. Use of powerful AI technology and also the NVIDIA Mellanox HDR InfiniBand smart switches will allow researchers of association quickly and to effectively carry out any tasks of an exaFLOPS order. Jan Buck, the CEO and the vice president of division of intensive calculations of NVIDIA company commented |
Upgrade of infrastructure will begin in February, 2021. It is expected that the first phase of the supercomputer will be set and ready to work in the middle of 2021.