Supercomputers are used for heavy processing. Supercomputer term is usually used for systems which have a high data processing power in few seconds. Wikipedia defines supercomputers as:
The computer that is applied at the time of its introduction in the computing capacity per unit of time is leading in the world. The word of supercomputer was first used by New York World magazine to refer to IBM at Columbia University.
These computers are primarily used for scientific and engineering projects which require very fast and complex calculations.
Some of the things that supercomputers are involved in their producing process are:
- Experiment mathematical models for complex physical phenomena
- Future climate simulation
- Simulate the genesis and evolution of the universe
- Nuclear weapons simulation and reactor operation
- Research on chemicals to make new medicines
- Cryptography
When the cost of supercomputers in 1990s declined, more businesses began to use supercomputers for researches and other business-related projects.
Supercomputer Features
Supercomputers have distinctive features. Unlike ordinary computers, supercomputers usually have more than one CPU. These processors have special circuits for interpreting program information and instructions and perform practical and logical calculations in the appropriate sequence.
It is essential to use multiple powerful processors in heavy and complex computing. Electronic signals cannot travel faster than the speed of light. Therefore, speed limits are the main problem for signal transmission and switching CPU circuits.
Achieving high computational speed in CPUs needs a fast data and instructions recovery. Hence, most of the supercomputers use very large storage capacities. Supercomputers also have very fast input and output capabilities.
Another distinguished feature of supercomputers is their vector arithmetic for processing which allows them to process several calculations at the same time.
For example, a typical supercomputer can multiply the list of hourly wages for a group of factory workers by the list of working hours of the members of that group to get a list of the profit earned by each worker.
Applications of Supercomputers
Supercomputers were originally used in national security applications, including nuclear weapons design and cryptography. Today they are commonly used in the aerospace, oil and automotive industries. In addition, supercomputers have found many applications in engineering or scientific research.
Undoubtedly, supercomputers have played a major role in our understanding of the universe, nuclear weapons design, and cryptography. Supercomputers also have many applications in weather forecasting. These predictions are based on numerical models.
By reducing the cost of building supercomputers and the increasing development of technology and providing a suitable platform for online games, the use of these computers in today`s industry became as usual.
In 2007, World of Warcraft supercomputers were able to gather together more than a million people in the online gaming world and open the way of progress in this area.
Unit of Measurement in Supercomputers
TFLOP is short form of “teraflop”. Teraflop is a method for calculating computer hardware power based on mathematical capacity. One teraflop is the capacity of a processor to calculate one trillion floating points per second.
For example, when we say the power of a piece of hardware is 6 teraflops, it means that it can calculate and process 6 trillion floating points per second.
What is a floating point calculation?
Calculating floating points is a common way to measure the systems processing power. In fact, as we talk about teraflops, we will have a universal and standard unit for calculating the power of computers.
If you recall your math lesson, floating points are actually “real numbers”, which from a mathematical point of view in any kind of finite calculations that use real numbers (especially in the decimal system).
Generally, teraflop is the number of mathematical equations based on real numbers that processor is able to process them per second. Different devices may require different amounts of flops. For example, a typical calculator with only 10 flops can do all your math commands and calculations.
So when we talk about megaflops (million flops), gigaflops (billion flops) and teraflops (trillion flops), you can imagine how much power we are talking about.
Manufacturers of different hardware typically use flops to express the processing speed of their devices.
Each petaflop is a teraflop or a quadrillion (number one with 15 zeros to the power of two) units of velocity.
Top 10 Supercomputers in the World
According to the latest ranking of the fastest supercomputers, which was recently announced, the top ten supercomputers in this list were also determined, we will look at these supercomputers in this article.
10- Lassen
Lassen, Sierra’s little brother at Lawrence Livermore National Laboratory, built by IBM with the same architecture. Its recent improvements have increased its speed to 18.2 petaflops.
Unlike its brother, Lassen is dedicated to unclassified research. This supercomputer was made of 795 nodes, 34,848 CPUs and 3,168 NVIDIA graphics (GPUs), and has 253,440 RAMs.
9- SUPERMUC-NG
In 2018, the new generation of SuperMuc supercomputer was officially launched at Leibniz Supercomputers in Garching (near Munich, Germany) and started working in 2019. It is the most powerful supercomputer in the European Union which is built with Lenovo and Intel`s technology. SuperMUC is the most powerful supercomputer that powered by only x86 series processors. The processing speed of the SUPERMUC-NG supercomputer reaches 19.5 petaflops.
8- ABCI Supercomputer
It is a system built by Fujitsu and belonging to Japan’s National Institute of Advanced Industrial Science and Technology, this machine places on eighth grade in the ranking.
However, this supercomputer has dropped one grade compared to the previous ranking. ABCI’s was built to serve as a cloud-based Artificial Intelligence resource available to Japanese companies and research groups.
7- Trinity
The TRINITY supercomputer is one of the most powerful computers in the world, manufactured by Cray by the Los Alamos National Laboratory, and ranked seventh in the top supercomputers.
This machine, which inherited its name from the first U.S. nuclear test in 1945, is mainly devoted to nuclear weapons-related calculations. The world’s first atomic bomb was built at this lab under the direction of Dr. Robert Oppenheimer during the Manhattan Project.
The processing power of the TRINITY supercomputer reaches 3.678 (GFl/W)
.
6- Piz Daint
Europe’s most powerful system ranks sixth on the list. This supercomputer is located at the Swiss National Supercomputing Centre and it named after the Piz Daint mountain in the Alps.
By the end of 2016, the Piz Daint supercomputer had reached to the record 25 petaflops, ranking third in the supercomputer at the time. It is an upgrade of a system built by the American company Cray, founded by the father of supercomputing Seymour Cray.
Its Intel and NVIDIA processors give it a speed of 21.2 petaflops. Piz Daint is involved in extensive research in materials science, physics, geophysics, life sciences, climatology and data science.
5- Frontera Supercomputer
The frontera supercomputer was developed by Dell Company and it is housed at the University of Texas Advanced Computing Center. This supercomputer is equipped with Intel processors. Frontera was able to quickly open its place among the bests in this field and reached the fifth rank.
Frontera was unveiled to the world in September 2019 as the world’s fastest supercomputer located in a university. Since June, it has been collaborating with three dozen scientific teams in research related to the physics of black holes, quantum mechanics, drug design and climate models.
Its 23.5 petaflops have been available to the scientific community, which benefit from its computational capacity especially in the areas of astrophysics, materials science, energy, genomics and the modelling of natural disasters.
4- Tianhe-2a
Chinese supercomputer TIANHE-2A or The Milky Way got the fourth place on this list. This supercomputer was developed by the National University of China and equipped with Intel Xeon processors with nearly 5 million cores that allow it to reach 61.4 petaflops.
According to Chinese officials, the machine is use for computing related to China`s government security.
3- SUNWAY TAIHULIGHT
China was ranked first in supercomputers with TAIHULIGHT supercomputer for 2 years but with the advent of two supercomputers, which you will read in the following article, it has fallen to third place.
However, this system is still one of the bests in this field. Taihulight was built by the National Centre for Engineering Research and Computing Technology and installed at the National Supercomputing Centre in Wuxi. Unlike other machines of its class, this supercomputer lacks accelerator chips and its processor speed reaches 93.0 petaflops.
TaihuLight is in a way a product of the war between China and the US, since its construction has completely dispensed with US technology, in response to the restrictions imposed by the US. This supercomputer has participated in research such as the Simulation of the birth and expansion of the universe using 10 billion digital particles.
2- Sierra
IBM is also responsible for the second most powerful supercomputer on the list, Sierra, located and built in California’s Lawrence Livermore National Laboratory.
This supercomputer has 1,572,480 cores powered by IBM Powe9 processors; also, powered by Nvidia Volta GV100 accelerators. Sierra manages 94.6 petaflops and it has been unchanged in the last 6 months ago.
1- Summit
Without a doubt, IBM is now the leader in supercomputers which summit, the most powerful supercomputer in the world, is made by this company for the U.S. Department of Energy’s Oak Ridge National Laboratory in Tennessee.
It occupies the equivalent of two basketball courts and achieves an impressive 148.6 petaflops thanks to its 2.41 million cores. Summit, eight times faster than America’s previous supercomputer, is a huge improvement for the US and even the world.
Oak Ridge team says this system which costed 200 million dollars to built is the first supercomputer that is made to use in artificial intelligence applications.
