The 10 Most Powerful Supercomputers in the World: Here’s What They Are and Where They Are Located

The world of supercomputer It is one of the keys to the advancement of scientific research and this is also why it is in constant evolutionwith machines that constantly challenge the limits of computation and performance and end up in the rankings of the most powerful computers in the world. The most powerful in absolute terms according to the latest rankings published in 2024 I am the American Frontier and the Japanese Fugakudepending on the classification method being considered.
Yes, because the rankings can be drawn up using two different methods. The first, HPL (High Performance Linpack), measures the speed at which a supercomputer performs dense linear algebra operations, such as multiplying large matrices, and is essential for tackling complex problems such as simulating climate change or finding new drugs. The second, HPCG (High Performance Conjugate Gradient), focuses instead on sparse linear algebra operations, evaluating the supercomputer’s ability to solve systems of linear equations with sparse data (i.e. data in which many elements are zero), and is crucial for applications such as traffic flow optimization or neural network simulation. Both methods are essential to address advanced scientific and technological challenges.
Regardless of the method used, one of the key factors to consider is the so-called computing power which is the measure of how many operations a certain calculator is able to do per second and is expressed in FLOPSor operations between numbers with a decimal point performed every second (FLoating point OPerations for Ssecond).

Who decides which are the most powerful supercomputers in the world?

These rankings are drawn up by Top500: a recognized point of reference for the technology and scientific community worldwide. This list is published twice a year (in June and November) and provides an up-to-date overview of the state of global supercomputers and represents an important resource for the scientific community, high-performance computing experts and companies operating in the sector. In reality, it would be more correct to actually speak of “lists” (plural) since Top500 publishes two biannually: one based on the HPL method and one based on the HPCG method.

The latest lists, published in June 2024see the US supercomputer Frontier HPL and Japanese supercomputer top the charts Fugaku at the top of the HPCG one. The only supercomputer hosted in Italy present in the top ten of both rankings is Leonardoranked seventh in the HPL rankings and sixth in the HPCG rankings. Below we offer you both top 10s in full.

The ranking according to the HPL method

1. Frontier: has 8.7 million cores which provide a computing power of approximately 1200 petaFLOPS/s (i.e. 1200 million billion operations per second) according to the HPL method. This nominal power makes Frontier the only supercomputer in the world belonging to the family of the so-called “exascale” computers, that is, those computers that are able to work at an order of magnitude of billions of operations per second. It is located in the DOE/SC/Oak Ridge National Laboratoryin the United States.
2. Aurora: has 9.2 million cores which provide a computing power of approximately 1000 petaFLOPS/s according to the HPL method. It is found in the DOE/SC/Argonne National Laboratoryin the United Statesthe.
3. Eagle: has 2 million cores which provide a computing power of approximately 561 petaFLOPS/s according to the HPL method. Microsoft’s supercomputer is located at Microsoft Azurein the United States.
4. Fugaku: has 7.6 million cores which provide a computing power equal to approximately 442 petaFLOPS/s according to the HPL method. It is located at the RIKEN Center for Computational Sciencein Japan.
5. LIGHTS: has 2.7 million cores which provide a computing power of approximately 531 petaFLOPS/s according to the HPL method. It is located at theEuroHPC/CSCin Finland.
6. Alps: has 1.3 million cores which provide a computing power of approximately 353 petaFLOPS/s according to the HPL method. It is located at the Swiss Supercomputing Center (CSCS)in Swiss.
7. Leonardo: has 1.8 million cores which provide a computing power of approximately 306 petaFLOPS/s according to the HPL method. It is located at theEuroHPC/CINECAin Italy.
8. MareNostrum 5 ACC: has 0.6 million cores which provide a computing power of approximately 175 petaFLOPS/s according to the HPL method. It is located at theEuroHPC/BSCin Spain.
9. Summit: has 2.4 million cores which provide a computing power of 148 petaFLOPS/s according to the HPL method. It is located at the DOE/SC/Oak Ridge National Laboratoryin the United States.
10. Eos NVIDIA DGX SuperPOD: has 0.49 million cores which provide a computing power of approximately 121 petaFLOPS/s according to the HPL method. It is located at the headquarters of NVIDIA Corporationin the United States.

The ranking according to the HPCG method

1. Fugaku: has 7.6 million cores which provide a computing power of approximately 16 petaFLOPS according to the HPCG method.
2. Frontier: has 8.7 million cores which provide a computing power of approximately 14 petaFLOPS/s according to the HPCG method.
3. Aurora: has 9.2 million cores which provide a computing power of approximately 5.6 petaFLOPS/s according to the HPCG method.
4. LIGHTS: has 2.7 million cores which provide a computing power of approximately 4.5 petaFLOPS/s according to the HPCG method.
5. Alps: has 1.3 million cores which provide a computing power of approximately 3.6 petaFLOPS/s according to the HPCG method.
6. Leonardo: has 1.8 million cores which provide a computing power of approximately 3.1 petaFLOPS/s according to the HPCG method.
7. Summit: has 2.4 million cores which provide a computing power of approximately 2.9 petaFLOPS/s according to the HPCG method.
8. Perlmutter: has 0.88 million cores which provide a computing power of approximately 1.9 petaFLOPS/s according to the HPCG method. It is located at the DOE/SC/LBNL/NERSCin the United States.
9. Sierra: has 1.5 million cores which provide a computing power of approximately 1.7 petaFLOPS/s according to the HPCG method. It is located at the DOE/NNSA/LLNLin the United States.
10. Selene: has 0.55 million cores which provide a computing power of approximately 1.6 petaFLOPS/s according to the HPCG method. It is located at the headquarters of NVIDIA Corporationin the United States.