4 supercomputers from Cyfronet among the world’s fastest and most energy-efficient computers
On the 18th of November, during The International Conference for High Performance Computing, Networking, Storage, and Analysis in Atlanta, USA, new lists of the TOP500 and Green500 - the world's fastest and most energy-efficient supercomputers - were announced.
Helios supercomputer, which works in Cyfronet, was recognised among the world's top 100 fastest and top 10 most energy-efficient computing machines! Thus, it maintained its leading position in both these categories in Poland. In addition, for the second time in the history of Polish computer science, as many as 4 places on the prestigious TOP500 list were occupied by supercomputers from one centre - Cyfronet.
Helios – new light for innovation in Polish science and economy
Helios has 35 PFLOPS of theoretical computing power, more than 108,000 computing cores, 460 TB of RAM, and 18 PB of disk system capacity, which together offer performance of nearly 2 TB/s. The supercomputer was built according to Cyfronet's design by Hewlett-Packard Enterprise based on the HPE Cray EX4000 platform and consists of three computing partitions:
- CPU equipped with 75,264 AMD Zen4 computing cores and 200 TB of DDR5 RAM,
- GPU equipped with 440 NVIDIA Grace Hopper GH200 Superchips, and
- INT for interactive work, equipped with 24 NVIDIA H100 accelerators and high-speed NVMe local storage.
A 200 Gbps Slingshot network interconnects all supercomputer components. The HPE Cray EX4000 platform Helios is built on is also used to construct the fastest supercomputers in the world (El Capitan, Frontier). The system's CPU and GPU partitions are liquid-cooled, which makes it possible to achieve a very low PUE (Power Usage Effectiveness), resulting in increased energy efficiency and lower operating costs. In addition, the heat recovered in this way can be used for heating.
On the current TOP500, Helios is listed twice. Its GPU partition ranked 69th, and its CPU partition 348th.
The division into two separate positions is due to the desire to showcase the full capabilities of each computing partition by using specially optimized versions of the benchmarks created as a result of Cyfronet's experts' cooperation with HPE, NVIDIA and AMD.
Helios' computing power for AI calculations is 1.8 EXAFLOPS!
Helios carries out many research projects for scientists from all over Poland, including in chemistry, physics, materials engineering, energy, medicine, biology and artificial intelligence. Helios was built as part of a project under the Intelligent Development Operational Program, so its computing power can also be used commercially by the economy.
Athena and Ares
Athena supercomputer achieves the theoretical computing power of over 7.7 PetaFlops (7,709 TeraFlops), giving the machine 212nd place in the November TOP500 list and making it the fifth fastest supercomputer in Poland.
Athena's configuration includes 48 servers with AMD EPYC processors (6,144 CPU cores) and 384 NVIDIA A100 GPGPU cards. The computer’s architecture meets the needs of users, who use the computing infrastructure both to perform standard high-performance scientific simulations and to apply artificial intelligence (AI) and machine learning (ML) methods to research in medicine, pharmacology, biology, chemistry, physics and many other scientific fields.
Athena’s computing power for AI calculations is almost 240 PFLOPS!
Ares supercomputer offers a total computing power of over 4 PetaFlops (the theoretical performance of the CPU part is over 3.5 PetaFlops, and the GPU part is over 0.5 PetaFlops). This power comes from computing servers with Intel Xeon Platinum processors (37,824 computing cores) and 72 NVIDIA Tesla V100 cards. On the November TOP500 list, Ares ranked 490th.
World-leading energy efficiency
A huge success is the receipt of the measurement of the energy efficiency of Helios' GPU partition, which placed it seventh on the Green500 list.
Energy efficiency is measured in terms of the amount of energy required to maintain a unit of computing power. Such a high ranking means that Helios offers more computing power for every kilowatt hour consumed than less efficient systems, so computing using it is cheaper than other machines and has less environmental impact.
At the same time, Athena ranked 53th on the Green500 list, showing excellent energy efficiency, especially considering that this supercomputer is air-cooled, not liquid-cooled. Cyfronet's other two supercomputers on the Green500 list also ranked high: Helios CPU was ranked 113, and Ares was ranked 145.
Polish supercomputers on TOP500 and Green500
The TOP500 list included 8 supercomputers from Poland (theoretical computing power in parentheses):
# 69 - Helios GPU (30.44 PFLOPS) - ACC Cyfronet AGH
# 94 - Proxima (23.32 PFLOPS) - PCSS IChB PAN
# 100 - Lem (20.37 PFLOPS) - WCSS PWr
# 186 - Kraken-Fregata (10.02 PFLOPS) - CI TASK PG
# 212 - Athena (7.71 PFLOPS) - ACC Cyfronet AGH
# 292 - Altair (5.88 PFLOPS) - PCSS IChB PAN
# 348 - Helios CPU (3.35 PFLOPS) - ACC Cyfronet AGH
# 490 - Ares (3.51 PFLOPS) - ACC Cyfronet AGH
As much as half of them are machines working at ACC Cyfronet AGH!
The supercomputers listed on the TOP500 are also included in the Green500 ranking:
# 7 - Helios GPU - ACC Cyfronet AGH
# 53 - Athena - ACC Cyfronet AGH
# 113 - Helios CPU - ACC Cyfronet AGH
# 145 - Ares - ACC Cyfronet AGH
# 160 - Altair - PSNC IBCh PAS
# 248 - Proxima - PSNC IBCh PAS
# 251 - Lem – WCNS WUST
# 289 - Kraken-Fregata - CI TASK GUT
The World’s and European winners
The world's fastest supercomputer has been declared El Capitan, installed at Lawrence Livermore National Laboratory Oak in the US. El Capitan was built on the HPE Cray EX platform, just like the Polish Helios, with more than 11 million computing cores, which secured its position as the world leader. The TOP10 list included 5 machines from the US, 4 from Europe and 1 from Japan.
The fastest European supercomputer was the HPC6 supercomputer installed in Italy. Three places lower was the LUMI supercomputer, which Polish scientists can access through the PLGrid infrastructure. ACC Cyfronet AGH coordinates activities in this area. For more information, visit: https://cyfronet.pl/lumi.
Use the power of supercomputers in your research
Supercomputers make it possible to significantly reduce the time of performing calculations, which using single computers would often take many years.
Cyfronet’s users benefit from professional support – from training sessions to individual expert consultations. Information about Cyfronet's resources is available at https://cyfronet.pl. The list of services within the PLGrid infrastructure is available in the PLGrid website.