The 5 top powerful supercomputers in the world

With the passage of time, supercomputers constantly advance. Once in a while, newly built supercomputers break the world record in terms of performance and pushing the limits of speed. Today, supercomputers have enabled us to go beyond our limits. Artificial intellegence (AI) development, scientific modellings and discoveries, engineering enhanced simulations, etc. as some achievements of high-performance computing (HPC) have empowered us to be smarter. Generally, advances in many areas in recent years have been taken place because we had built HPC systems that are powerful enough to process the work.

The processing power of supercomputers are usually stated by two values of Rmax and Rpeak. Rmax is the maximum computing performance achieved on LINPACK benchmark (HPL) which is a measure of computing power. Rpeak is the theoretical maximum performance obtained by the supercomputer hardware. To report a supercomputer processing speed, Rmax, which is the practical processing speed, is used.

In this essay, the 5 top supercomputers are introduced by 2024:

1. Frontier

Frontier supercomputer

Currently, Frontier is the most powerful and the first exascale supercomputer in the world. It was developed with a budget of about US$ 600 million at the Oak Ridge National Laboratory in Tennessee, United States. In August 2022, it began operating with performance of 1.2 exaFLOPS on LINPACK benchmarks (Rmax). In other words, it can do 1.2 quintillion floating operations per second! Frontier is based on HPE Cray’s new EX architecture and has about 9500 64-core AMD EPYC CPUs and 38000 AMD Instinct MI250X GPUs. Scientists originally planned to use Frontier for drug discovery, cancer research, unknown materials, nuclear fusion, and modeling star explosions. It is reported that scientists will use Frontier to design new transportation and medical technologies.

2. Aurora

Aurora supercomputer

Designed by Intel and Cray at Argonne National Laboratory in Illinois, United States, Aurora became the second exascale supercomputer in the world. It cost US$500 million and was open for early operations in June 2023. Basically, the purpose of Aurora is scientific development and research such as subatomic particles, energy storage and fusion, cancer, low carbon technologies, etc. Aurora has more than 21000 Intel Xeon Max Series CPUs and 63000 Intel Data Center GPUs working in total racks of 166 which reach the Rmax performance to 1.012 exaFLOPS. It is believed that this super-powerful computing system has a significant role in further nuclear fusion achievements. 

3. Eagle

Eagle supercomputer

Eagle supercomputer is not located in a laboratory — it is actually in the cloud, and anyone can access it via Microsoft’s Azure cloud platform. It is a distributed network of systems that together have enough power to become the third fastest in the top 500 supercomputers. Its processing speed reaches 560 petaFLOPS. Eagle was made available to the public in August 2023 and in theory, is accessible by anyone willing to pay. The CPUs are 8480 Intel Xeon Platinum C and the GPUs are 100 Nvidia H.

4. Fugaku

Fugaku supercomputer

Fugaku was built in June 2020 (US$1billion) and for two years, it was the fastest supercomputer in the world. Fugaku is based at Riken Computational Science Center – Kobe, Japan. The name Fugaku is derived from Mount Fuji, a volcano about 100 kilometers from Tokyo. Over the years, researchers have used fugaku to answer critical research questions. During the Covid-19 pandemic, they undersrtood that non-woven face masks were more effective at clogging airborne respiratory droplets. This achievement was obtained by tremendous number-crunching capabilities to process data. The Japanese say at present, Fugaku is training Japanese AI language models in the form of ChatGPT. Fugaku is composed of approximately 160000 Fujitsu A64FX microprocessors which enables the supercomputer get the maximim speed of 442 petaFLOPS.

5. Lumi

Lumi supercomputer

Lumi with the budget of US$ 153.5 million started to work in June 2021 at CSC Data Center – Kajaani, Finland. It has the rank 5 in the TOP500 List with performance of 380 petaFLOPS. It is equipped with 3rd generation EPYC 64-core CPUs (totally 362000 coes) and AMD Instinct MI250X GPUs (145 million cores). Lumi is considered the most powerful supercomputer in Europe. According to European Union, it uses 100% renewable hydroelectric energy and its lost heat is reused to heat the buildings around it.

Final words

It seems that supercomputers are incredibly promising for more complex tasks in the future. With constant advances, they are likely to have even more wide-ranging impacts across technology, science, and industry. In the coming years, the top 5 supercomputers are expected to lose their ranks and be replaced by other newly built supercomputers.

Related links