A supercomputer is a computer that has a higher level of performance than a general-purpose or personal computer and even workstation and server computers. It does things a thousand times better than other computers of its generation. The demand for machines that can handle high computational workloads at a faster rate and the applications of high-performance computing have remained the basis for the invention and further development of supercomputing. Hence, because of their over-the-top and unparalleled level of performance, supercomputers are also considered the titans of the computing world. What exactly makes it “super” when compared to other computers? What are the characteristics of a supercomputer that set it apart from personal or general-purpose computers and enterprise-grade computers?
Understanding Supercomputers: The Main Characteristics of a Supercomputer
One of the first supercomputers was the Livermore Atomic Research Computer. It was invented in 1960 by the Eckert-Mauchly Computer Corporation. The introduction of the CDC 6600 which was designed by American electrical engineer Seymour Cray and developed by Control Data Corporation in 1964 marked the transition from germanium to silicon transistors.
Take note that the CDC 6600 also became the first computer to be called a supercomputer due to the fact that outperformed other computers of its time. However, after Cray left Control Data Corporation to establish Cray Research, he developed the Cray-1 in 1976. It became one of the most successful supercomputers in history due to its wider market reception.
Some of the supercomputers developed in the past are still operational. More modern ones are better on all fronts compared to the older-generation ones. The advancements in hardware and more specific chipmaking technologies have pushed further the capabilities of these machines and expanded their applications while allowing them to become more accessible.
Nevertheless, based on their name alone, the main characteristic of supercomputers is their high floating-point operations per second or FLOPS. There are machines that can perform over 100 petaflops. Note that the performance of personal computers ranges from hundreds of gigaflops to tens of teraflops. Below are the more specific characteristics of a supercomputer:
1. Powered By Multiple Central Processing Units and Dedicated Processors
A typical general-purpose computer has a single central processing unit. A supercomputer has more than one. Some even have hundreds to thousands of cores or smaller processing units. A multiple CPU setup enables the distribution of processing power to tackle multiple data and instructions or complex tasks in parallel and achieve high computational speed.
The overall hardware architecture of a supercomputer can also be similar to a modern personal computer in the sense that it uses other processors or coprocessors. These include graphics processing units, field-programmable gate arrays, artificial intelligence accelerators, and data processing units. The architecture depends on the specified use case of the machine.
2. Requires Specialized Software Written in Parallel Programming Languages
Another characteristic of a supercomputer is that it requires specialized software programs written in parallel programming languages. A parallel programming language supports the creation and execution of multiple tasks or instructions at the same time on one or more processors. It can improve the performance, efficiency, and scalability of software applications.
Hence, in supercomputing, to maximize the capabilities of multiple processors, supercomputers use specialized software for parallel operations. This allows tasks or instructions to be broken down and distributed across multiple cores or processors. The result is a faster and more efficient processing of large datasets, intensive and higher-order computations, and simulations.
3. Uses Multiple and Massive Random Access Memory and Storage Devices
Supercomputers also use massive RAM measured in terabytes or even petabytes. It also does not use a single RAM device. The overall RAM is based on a distributed architecture that involves thousands of nodes. Each node has its own memory module. These enable the handling of large datasets, generation of intermediate results, and parallel data access and processing.
The storage capacity is also large or between the terabytes and petabytes territories. Some design uses a combination of high-performance hard disk drives, solid-state drives, or magnetic tapes. The colossal and multiple storage space of supercomputers allow them to store and retrieve large amounts of data for on-demand processing and during data generation processes.
4. High-Speed Input and Output Operations and Communication Processes
Another characteristic of a supercomputer is its fast input and output operations. It can interface and communicate with hardware components, other devices, and networks at a high speed for a lag-free signal transmission and smoother overall operations. Examples of these components and devices are storage mediums, multiple processors, memories, and network interfaces.
It is important to note that the sheer volume of data that moves within a supercomputer requires super-fast interfacing channels. There are high throughput networks for connecting components and devices. Both the communication interfaces and the components or devices themselves are also high-performance. These are critical in maximizing computational capabilities.
5. Requires Advanced and Complex Heat Management and Cooling System
The operations of multiple processors and high-performance components and devices at faster rates generate immense amounts of heat. The IBM Summit produces 44.5 megawatts of heat. This is equivalent to 14500 average households in the United States. The Tianhe-2 produces about 60 megawatts of heat and the Sunway TaihuLight produces 51 megawatts of heat.
Nevertheless, because of the large amount of heat they generate, these supercomputers require the use of advanced and complex heat management and cooling systems. The IBM Summit is equipped with a liquid cooling system that circulates 4000 gallons of water per minute through internal pipes and dedicated heat exchangers to remove and dispel excess heat further.
6. Colossal Physical Size that Requires Housing in Large Physical Real Estate
It is important to note that supercomputers do not look like general-purpose desktop computers. Their massive size is another one of their defining characteristics. This comes from a modular orientation. An entire room is populated with different cabinets that house specific components and devices. The cooling system is also built within the room and around the cabinets.
Fugaku in Japan occupies a floor space of about 3300 square meters and Sunway TaihuLight in China occupies a floor space of about 1000 square meters. These supercomputers are two of the fastest supercomputers in the world. Nevertheless, based on these, one can consider facilities or specific room or floor spaces as the chassis or enclosures of supercomputers.
Summary and Takeaways: Characteristics of a Supercomputer in a Nutshell
Remember that the defining characteristic of supercomputers is that they are thousands of times more powerful than a personal computer. These machines are designed for high-performance parallel processing. They are also massive. These features also make them power-hungry heat generators that require specialized heat management and cooling systems.
It is also worth mentioning that not all supercomputers are built the same. Their core function is within the realms of computational science, physical simulation, or cryptanalysis. However, because each of these realms has different subfields or applications, these machines also often feature specialized architectures optimized for specific types of calculations.