Top 10 Most Powerful Supercomputers In The World

Supercomputers are a great unknown, even though many of us make use of them daily without knowing it, such as when we use certain cloud services. These teams are spread across developed countries and there is a list that brings together the most powerful in the world called Top500. In this article, we bring you closer to these great machines and we show you the current Top 10 of these great machines.

What is HPC?

What is HPC?
What is HPC?

High-Performance Computing (HPC) is the ability to process data and perform complex calculations at high speed. High-Performance Computing (HPC) achieves these goals by aggregating processing power so that even complex applications are efficient, reliable, and fast, according to the needs and expectations of users.

By bringing diverse technologies such as computer architecture, algorithms, systems software, programs, and electronics under one umbrella, HPC can solve complex problems quickly and efficiently by operating as a single system. Its true power comes from connecting multiple HPC nodes in a cluster.

What are supercomputers?

What are supercomputers?
What are supercomputers?

Supercomputers are large machines with characteristics far superior to those we have at home. And they achieve this thanks to the redundancy of elements such as the CPU, the GPU, the memory units, etc. For example, a supercomputer might have thousands or hundreds of thousands of processors to be able to reach the computing powers of several petaFLOPS.

However, the elements it has are quite similar to those of a conventional team. For example, storage units or RAM memory modules are quite similar. Some differences are usually in the motherboards, which are usually MP or MultiProcessor, that is, with several sockets, in addition to CPUs of specific models for HPC such as Intel Xeon or AMD EPYC, among others such as IBM POWER, SPARC, Arm, etc.

For all the units to work in unison, distributing the workload among all of them and thus achieving those performance rates, a multitude of network units such as switches are used, which intertwine all these elements that would work as a single PC of epic proportions.

Something similar happens with the GPUs of these computers, which are also usually special versions for HPC (High-Performance Computing), such as those from NVIDIA Ampere or the AMD Radeon Instinct, specially enhanced to obtain a higher FLOPS rate, or optimized for tasks such as AI, the ML, etc.

On the other hand, storage units are usually HDD or SSD very similar to those we have in a conventional PC, but they are usually installed redundantly to achieve higher capacities and with RAID configurations for greater reliability if any of the units fail so as not to lose the data. However, some brands such as Seagate or Western Digital have specific models of their special storage units for high-availability machines.

As for the brands of this equipment, they are also quite well known to all of you. For example, HPE (HP) is one of the largest builders of supercomputers, along with Fujitsu, IBM, and Lenovo, among others.

And I would not like to end this section without also talking about the software. 100% of the world’s most powerful supercomputers use some version of Linux, especially SUSE Linux Enterprise Server (SLES) or RHEL (Red Hat Enterprise Linux).

Refrigeration systems

Given the dimensions of this equipment, which can weigh tons and measure hundreds of square meters, all this hardware working with high loads releases a lot of heat and needs to be cooled to work properly. Therefore, conventional cooling systems are not used as in a PC.

For these systems, refrigeration systems based on air conditioning for the room and also other advanced systems such as liquid cooling, and even some by immersion in a special liquid that cools this equipment to ideal temperatures, are usually used.

On the other hand, not only the temperature is controlled, but also the relative humidity of the room where these machines are housed (data center) and other parameters, since they can be very susceptible to variations and humidity and short circuits that drop the service.

Feeding systems

It is also important to highlight the power system of these teams. They usually have more powerful power supplies that in their entirety can be feeding hundreds of MW (megawatts) to the machine, since these HPC equipment consume huge amounts of energy. Data centers consume more than 1% of the world’s energy.

In addition, they are complemented by UPS or uninterruptible power supply systems so that the centers continue to function in the event of a power outage. These UPS systems usually last about 15 minutes, enough for other diesel generator units that usually have these centers to power the entire system when the power outage lasts longer than that. Keep in mind that they are equipment created for high availability, and all the time they are stopped is a huge waste of money.

What are supercomputers used for?

What are supercomputers used for?
What are supercomputers used for?

This equipment costs billions of dollars and is often used for a wide variety of tasks. They can work to provide a cloud service to connected clients or for scientific tasks in many fields, from physics or medicine, to astronomy, biology, or environmental studies. For example, thanks to these machines, the COVID vaccine and the decryption of its DNA sequence were deciphered in record time.

These large machines are also used for nuclear research, the design of new technologies, or motorsport competitions. For example, Formula 1 teams have supercomputers to perform the CFD simulations necessary for the development of aerodynamics. Also aircraft manufacturers, etc.

They are also used for meteorological studies, with the information from the satellites they can process it to determine the forecast that we can all see on our mobile devices or TV. The truth is that these machines are used for more things than many believe and, despite that, they are still unknown.

Common uses

Among the most common uses of supercomputing are:

  • Military research, such as for the development of nuclear weapons.
  • Astrophysics studies.
  • Environment and sustainability, such as research on climate change.
  • Meteorology for forecasts and prevention of natural catastrophes.
  • Medicine and health, to obtain new medicines, vaccines, the study of diseases, DNA sequencing, etc.
  • Studies of atmospheric composition and air quality.
  • Big data for the analysis of huge amounts of data in record time.
  • Biomechanics for simulation.
  • For the energy industry.
  • Geophysical investigations.
  • Bioinformatics.
  • Cloud Computing.
  • Social simulations and anthropological studies.
  • Research on outer space.
  • CFD simulation for ducts, aerodynamics, fluid behavior, etc.
  • Others.

Top 10 most powerful supercomputers today

Top 10 most powerful supercomputers today
Top 10 most powerful supercomputers today

Finally, it is important to know which are the 10 most advanced and powerful supercomputers today. The list updated during November 2022 is:

1º Frontier

The Frontier is the most powerful supercomputer today. It is found in the United States and is used by the DOE/SC/Oak Ridge National Laboratory for environmental studies. This equipment has 8,730,112 computing cores and can reach 1,102 PTFLOPS (floating point operations per second). In addition, this equipment has a consumption of no less than 21,100 kW.

It consists of a team designed by HPE, Cray EX235a, with 3rd Gen AMD EPYC processors with 64 cores each and a clock frequency of 2 GHz, along with AMD Instinct MI250X GPGPU units.

2nd Fugaku

Fugaku is a very special supercomputer, as it was the first ARM-based supercomputer to rank 1 for performance. Currently, it has fallen to the second position. It is located in Japan and has been created by Fujitsu and RIKEN Center for Computational Science.

It has 7,630,848 ARM cores distributed by A64FX units with 48 cores each at 2.2 GHz. The consumption of this equipment is 29,899 kW, and it can achieve a computing power of 442.01 PFLOPS.

3rd LUMI

Another system built by HPE, the Cray EX235a, with AMD EPYC 64 core 2 GHz units along with AMD Instinct MI250X graphics cards. This equipment is currently the most powerful in Europe and is hosted in Finland, and is shared for scientific studies in various Nordic countries.

It has 2,220,288 processing cores and consumes 6,016 kW.

4th Leonardo

This equipment has been created by Atos for EuroHPC/CINECA. It has 1,463,616 processing cores and consumes 5,610 kW of power. Its calculation performance reaches 174.70 PFLOPS.

It’s a BullSequana XH200 system with 32-core Intel Xeon Platinum 8358 processors each running at 2.6Ghz and NVIDIA Ampere A100 SM4 graphics cards with 64 GB VRAM, along with a QuadRail NVIDIA HDR100 Infiniband network.

5th Summit

At the time this supercomputer would also occupy the first position in the Top500, but it has now dropped to fifth. This is a piece of equipment created by IBM for the DOE/SC/Oak Ridge National Laboratory in the United States.

It has 2,414,592 processing core units, with a total power of 148.60 PFLOPS and a consumption of 10,096 kW. It is based on the Power System AC922 system with IBM POWER9 22-core 3.07 GHz processors, and NVIDIA Volta GV100 cards, with Mellanox EDR Infiniband network.

6th Sierra

This system is also an IBM, based on the Power System AC922. In this case also for the DOE/NNSA/LLNL of the United States. It has POWER9 processors with 22 cores each at 3.1 GHz, NVIDIA Volta GV100 cards, and Mellanox EDR Ifiniband network.

As for processing cores, there are a total of 1,572,480, with a computing power of 94.64 PFLOPS and a consumption of 7,438 kW.

7th Sunway Taihu Light

Created by Sunway MPP and based on Sunway SW26010 processors with 260 cores each at a frequency of 1.45 GHz, it is a supercomputer for the National Supercomputing Center in Wuxi, China. It is intended for various investigations.

It has 10,649,600 processing cores and develops 93.01 PFLOPS of computing performance. In addition, its consumption is 15,371 kW.

8th Perlmutter

The Perlmutter is an HPE Cray EX235n system for the US DOE/SC/LBNL/NERSC. A piece of equipment for scientific research with 761,856 processing cores capable of reaching 70.87 PFLOPS and consumption of 2,589 kW.

It is based on AMD EPYC 7763 processing units with 64 cores each and a clock frequency of 2.45 GHz. It is also accompanied by NVIDIA Ampere A100 SXM4 cards with 40 GB of VRAM each.

9th Selene

This supercomputer has been designed by NVIDIA and is in the United States. It can reach 63.46 PFLOPS of performance and a consumption of 2,646 kW. It has 555,520 AMD EPYC 7742 processing cores with 64 cores each at 2.25 GHz and NVIDIA Ampere A100 cards, along with a Mellanox HDR Infiniband network.

10th Tianhe-2A

This Chinese supercomputer also came to occupy the first position in the Top500. This is a piece of equipment for the Guangzhou National Super Computer Center dedicated to multiple studies. This TH-IVB-FEP unit is based on Intel Xeon E5-2992v2 clusters with 12 cores each and a clock speed of 2.2 GHz. It also has TH Express-2 and Matrix-2000 technology. With 4,981,760 processing cores, it can reach 61.44 PFLOPS with a consumption of 18,482 kW.

BONUSES – Marenostrum

In Spain we not only have the one that was chosen as the most beautiful of all supercomputers, the BSC Marenostrum in Barcelona, ​​but it is also one of the most powerful in Europe and the world. It is used for various investigations, from the development of medicines, cancer research, DNA, climate, and engineering, to the development of new European processors.

Now a new update will arrive, the Marenostrum 5 (built by Atos and with European processors). This should place this Spanish team among the Top 10 of the list. The update is expected to reach 314 PFLOPS of maximum performance, which means 314 trillion floating point calculations per second. It will also have more than 200 PB of storage and 400 PB of active files.

For this, 151.41 million euros have been allocated, but until it arrives we have to settle for the Marenostrum 4. This 2019 version has 14 PB of storage capacity, Omnipath high-speed network, Intel Xeon Platinum processors with 24 cores each adding a total of 165,888 cores, a consumption of 1.3 MW per year, and a performance of 11.15 PFLOPS.

On the other hand, it has other clusters made up of heterogeneous architectures, such as IBM POWER9 processors, NVIDIA Volta GPUs, AMD EPYC Rome and AMD Radeon Instinct MI50 processors, and even ARMv8-based processors.

The Marenostrum 4 equipment has been designed by Lenovo, and currently occupies the 88th position of the To p500 in the world.

Future: Quantum computing

DNS Server
DNS Server

Quantum computers are already supercomputers themselves. They are based on the quantum bit, or qubit, something analogous to bits in classical computing. Quantum computers use quantum bits, or qubits, which handle information in very different ways. Classic computers, including smartphones and laptops, encode information in binary bits, or bits, which can be either 0s or 1s.

Instead, the laws of quantum mechanics allow qubits to encode exponentially more information than bits. These qubits can be in one or the other state or both, being able to increase the power of parallelism of these machines.

Quantum and conventional computing are two parallel worlds, with some similarities but also many differences. For now, although there are already several of these quantum computers in operation, created by companies such as IBM, Intel, Google, and D-Wave Systems, among others, they still have a lot of development ahead of them to be truly functional. At the moment, they are only used experimentally and as services (QaaS or Quantum as a Service) as is the case with IBM Q.

No one has shown a better way to make a fault-tolerant quantum computer, and various companies and research groups are investigating various types of qubits. Given the current pace of development, it is likely that current supercomputers based on the current paradigm still have a long way to go.

Also, read about How To Build Your Own Computer? Quick Guide

Website | + posts

Abram left his e-business studies to devote himself to his entrepreneurial projects. In 2017, he created the company Inbound Media and wrote articles about high-tech products for his Chromebookeur site. In 2019, Chromebookeur was renamed Macbound and became a general purchasing advice site. Today, Abram manages the development and growth of Macbound, surrounded by a young and talented team.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *