A supercomputer is a very powerful computer that can do billions of operations per second. They are computers that are very fast, large in size, expensive, and with the biggest storage capacity in the world. These computers have a large main memory (RAM) capacity that can reach a petabyte(PB) and operation speeds of exaFLOPS. Supercomputers differ from personal computers in a number of ways as indicated in another of your article here.
Supercomputers can be classified into general and special-purpose computers. The general is further categorized into a vector, tightly connected cluster, and commodity computers. the main characteristics of these computers are high processing speed into exaFLOPS and high main memory. supercomputers are used for scientific research, medical inventions, simulation, and weather forecast among many others as discussed in the article.
Supercomputer’s performance in FLOPS
Supercomputers carry out billions of operations per second. To measure their performance they use Floating-Point Operations per second (FLOPS). A FLOPS is a unit for measuring how many floating operations a computer processor can do per second.
FLOPS are preferred when evaluating supercomputers since most supercomputers are used for scientific research which requires the accuracy of floating numbers. The fastest supercomputers are doing exaFLOPS performance.
Summary of supercomputer performance measurement unit.
|The number of FLOPS||Prefix assigned||Full name given|
Supercomputers classification and types
Supercomputers are classified into either special or general purposes.
Special purpose supercomputers
These are designed for a specific purpose and cannot be used for any other function. They are designed to tackle a specific problem. These systems spend their energy and resource to solve that specific assigned problem. An example of a special-purpose supercomputer is IBM Deep Blue which was designed to play the chess game.
General purpose supercomputers
These are computer systems that can be used for a variety of activities. They can be used to solve different problems at different times. General-purpose supercomputers are further classified into 3:
1. Tightly connected clusters of computers
These are groups of computers connected and work as a single unit to solve a common problem. There are 4 ways that clusters can be created to connect these computers which give us 4 types of clusters: Director-based clusters, Two-node clusters, Multi-node clusters, and Massively parallel clusters.
2. Vector processor
This is where the processor is capable of executing an entire array of data elements at the same time rather than working on one element at a time. This introduces a kind of parallelism where all elements in an array are operated same time. These supercomputer processors are arranged in arrays that can be able to operate on many data elements at the same time.
3. Commodity computers
This is made up of ordinary (common personal computers) that are connected with high bandwidth fast Local Area Network (LAN). These computers then implement parallel computing, where they work on a single task together.
Features and characteristics of supercomputer
- Have multiple processors: a supercomputer is made up of thousands or millions of Central Processing Units (CPUs) that have many cores within them. The fastest computer (Frontier) has over 8.7 million cores.
- They operate at a very high speed: supercomputers operate at a very high speed compared to other current computing devices.
- High-capacity main memory: the system is made up of many nodes and each has its memory addresses which can total up to petabytes of RAM. The frontier which is the fastest computer in the world has around 9.2 petabytes of memory. Other supercomputers have even a high capacity of RAM.
- They utilize parallel processing: this is a technique where many processors work together to solve a specific calculation simultaneously. Each processor work on a part of the calculation to help solve the problem as fast as possible.
- Most runs on customized Linux versions of the operating system. Linux-based operating systems are used since they are open source and they execute commands faster.
Uses of supercomputers
- Weather forecasting: the systems are used to read and predict how the weather will change over time. They can also be used to test in case of a tsunami or other natural catastrophe happening and what damage can they cause.
- Oil and gas exploration: supercomputers are used to explore new gas and oil area and try and approximate how many resources are on the ground.
- Simulation: the systems are used to carry out simulations of situations that can be dangerous if they are done in real life. Simulation helps to be prepared in case a situation like that ever happens.
- Medical research: Supercomputers were used during the COVID-19 pandemic to try to get the best methods to solve the problems. They are also used to learning more about diseases and coming up with new medicine for different problems.
- Big data mining: big data is the latest avenue of analyzing data which requires a lot of computing power. These computers are used to process and analyze data and come up with patterns for decision-making.
- Military and defense: military uses these computers to simulate situations such as nuclear war and other latest warfare equipment.
- molecular modeling
- Cloud computing: companies like GOOGLE and AMAZON who are big players in cloud computing services provision used supercomputers to run their systems.
Examples of 10 most powerful supercomputers in 2022
According to the Top500 list for June 2022 the following are some of the top 10 examples of the most powerful supercomputers in the world.
- Frontier is ranked 1st with 1.102 Exaflop/s using 8,730,112 cores.
- Fugaku, from Japan, comes second with 7,630,848 cores which achieve 442 Pflop/s.
- LUMI system, in Finland at number 3 with 151.9 Pflop/s and 1,110,144 cores.
- Summit, an IBM-built system at ORNL in Tennessee, USA, is listed as Number 4 with a performance of 148.8 Pflop/s and 2,414,592 cores.
- Sierra in the USA is ranked 5th with a performance of 94.6 Pflop/s with 1,572,480 cores.
- Sunway TaihuLight is a system developed in China that comes number 6 with the performance of 93 Pflop/s and 10,649,600 cores.
- Perlmutter at number 7 can do 64.6 Pflop/s using 761,856 cores.
- Selene in the USA achieves 63.4 Pflop/s with 555,520 cores.
- Tianhe-2A (Milky Way-2A) comes as number 9 with 61.4 Pflop/s and 4,981,760 cores.
- The last in the top 10 supercomputers in the world is the Adastra system installed at GENCI-CINES which achieves 46.1 Pflop/s using 319,072 cores.
Advantages of using supercomputers
- They have a very high processing speed: supercomputers can do billions of operations per second. They are so fast to the extent that the work that an ordinary personal computer can take 1 year they can perform in minutes.
- High storage capacity: they store a lot of data hence they have a large storage capacity
- Multiuser: supercomputers are multiuser computers which means many users can work on them simultaneously.
- They are used to solve complex numeric calculations very fast.
Disadvantages of supercomputers
- Consume a lot of power: these computers have many processors which means they require a lot of power to power them.
- They overheat: many CPUs in these systems generate a lot of heat hence overheating. Supercomputer requires special types of cooling system to deal with generated heat.
- Expensive: these computers are the most expensive computer in the world. Mostly they are owned by the government or big institutions and companies.
- They are huge: supercomputer occupies a very big physical location. They can fit within a whole building floor.
- They require specialists to maintain and operate.
What types of operating systems are used in supercomputer
Supercomputers are designed for customized operations and so they mostly used open-source types of operating systems. Most of them use Linux-based versions to meet the need that the system was designed for.
These operating systems are mostly command-based to ensure they don’t use a lot of system resources to increase their efficiency in operation.