Supercomputers are the peak of computing technology, capable of processing vast amounts of data and performing complex calculations at rapid speeds. These high-performance machines are used for a wide range of applications, from scientific research to national security simulations, climate modeling, industrial design, and artificial intelligence (AI).
The escalating demand for massive volumes of cloud-based computing power is bringing supercomputing technologies into hyperscale data centers. And the rapid adoption of AI will likely see more supercomputers being built worldwide.
For instance, the UK government has just announced plans to invest £900 million in a new supercomputer that will be one of the most powerful in Europe to drive UK research and innovation into the potential and safe use of AI technology.
As the UK aims to build one of the world’s first, large-scale, open AI supercomputers, it raises the question, where are the world’s fastest supercomputers, and how many are there?
Before delving into the numbers, it’s important to understand what sets a supercomputer apart from a high-performance computer.
Supercomputers are typically characterized by their exceptional processing power. Their speed is measured in floating-point operations per second (flops), which refers to how many arithmetic calculations they can perform. While there is no strict threshold that separates a supercomputer from a regular high-performance computing (HPC) system, supercomputers are generally capable of performing at speeds measured in teraFLOPs (trillions) and petaFLOPs (quadrillions).
Supercomputers are used for a wide range of applications by governments, research institutions, and industrial corporations. They range in size from super-large scale to small-scale devices. There are so many supercomputers today that it is difficult to pinpoint the exact number. In addition, some systems are not known publicly for national security, geopolitical or competitive reasons.
However, the TOP500 project ranks the world’s 500 most powerful supercomputers biannually based on their performance on the Linpack benchmark, which measures the speed at which a computer can solve a system of linear equations. This provides a standardized way to compare supercomputer capabilities.
Where Are the World’s Fastest Supercomputers?
The 500 fastest supercomputers are spread across 34 countries, of which 27 countries have more than one system. From November 2017 until November 2022, China had the highest number of supercomputers.
But as of June 2023, the US has resumed the top position on the Top500 list with 150 supercomputers. China has 134 and is followed by Germany, which has 36 supercomputers. Japan and France round out the top five, with the UK, Canada, Brazil, South Korea, and the Netherlands completing the top 10.
The US is also the leader in terms of performance, measured in maximal Linpack performance achieved (Rmax) at 2,400,757 teraFLOPS (TFLOPS). That is well ahead of China’s 465,824 TFLOPS.
The Frontier system at the Oak Ridge National Laboratory (ORNL) in Tennessee, US, is the only exascale machine reported with a High-Performance Linpack (HPL) exceeding one Exaflop per second (1 EFLOP/s).
The system is based on the HPE Cray EX235a architecture and is equipped with AMD EPYC 64C 2GHz processors and 8,699,904 total cores. Frontier increased its HPL from 1.02 EFLOP/s in November 2022 to 1.194 EFLOP/s in June, a 17% increase. Exascale was seen as only an aspirational goal just a few years ago, indicating the rapid pace of technology development.
The fastest supercomputer was previously the 442 PFLOP/s Fugaku system at the Riken Center for Computational Science (R-CCS) in Kobe, Japan, from June 2020 until June 2022. Fugaku is powered by Fujitsu’s 48-core A64FX system on chip (SoC), making it the first number one system to be powered by ARM processors.
The Summit supercomputer at ORNL held the top spot from June 2018 until November 2019 with a performance of 122.3 PFLOP/s and is now in the fifth position with an upgraded performance of 148.60 PFLOP/s. The Sunway TaihuLight supercomputer at the National Supercomputing Center in the city of Wuxi, China, was the fastest from June 2016 until November 2017 and now ranks seventh with a speed of 93.01 PFLOP/s.
China’s supercomputer development has been affected by US sanctions on its access to semiconductor equipment and other advanced technologies that can have military or intelligence applications.
However, there are reports that China has other advanced supercomputers that are not benchmarked but are evident from scientific results published in research papers.
Power Requirements Drive Energy Efficiency Push
Supercomputers and HPCs face soaring energy costs, particularly as demand for resources for AI, machine learning, and high-performance data analytics (HPDA) rises. New central processing units (CPUs) and GPUs consume more power than older models. Concerns around carbon (CO2) emissions and government regulations are further driving moves towards energy efficiency and sustainability.
Published alongside the Top500 list, the Green500 list sorts the world’s most powerful supercomputers by their energy efficiency. The two most efficient supercomputers are in the US, followed by systems in France, Australia, and Sweden. Supercomputers in Finland and Germany contribute to the top 10.
The HPL performance of the top systems shows that vast power capability does not have to come at the expense of efficient energy use. For instance, the LUMI supercomputer in Finland is the fastest supercomputer in Europe and the third fastest globally, with an Rmax of 309.10 PFLOP/s. The system, fully operational since December 2022, is the seventh most efficient on the Green500 list, with an efficiency of 51.382 GFLOPs per watt.
LUMI is fully powered by CO2-free hydroelectricity and uses natural water-cooling systems to cool down its processors. In addition, all the waste heat it produces is reused to provide local district heating in its home city of Kajaani. The Leonardo supercomputer in Italy is the fourth fastest in the world and the 15th most energy efficient. All six supercomputers in the European High-Performance Computing (EuroHPC) Joint Undertaking (JU) are water-cooled to improve their energy efficiency and aim to operate as sustainably as possible.
Meanwhile, the Frontier supercomputer in the US ranks as the sixth most energy efficient with a 52.59 GFLOPs/watt rating despite its HPL score of 1.194 EFLOP/s.
Supercomputers play a vital role in advancing science, technology, and innovation across various applications in science, academia, military, and industry. The number of supercomputers in the world is a dynamic figure that changes as technology advances and new machines are introduced.
As technology continues to evolve with rising demand for cloud computing capacity and AI, the number of supercomputers is expected to grow, further expanding the frontiers of what is possible in computational science and engineering.