Golden Recursion Inc. logoGolden Recursion Inc. logo
Advanced Search
Supercomputer

Supercomputer

Supercomputers refer to any of a class of high-level computers with performance far exceeding that of a general-purpose computer.

Overview

A supercomputer refers to any of a class of high-level computers with performance far exceeding that of a general-purpose computer. The phrase is generally used to describe the fastest and most powerful computers at any given time. Supercomputers are made up of interconnects, I/O systems, memory, and processor cores. Compared with traditional computers, supercomputers use multiple central processing units (CPU), that are grouped into compute nodes. Supercomputers can contain tens of thousands of nodes, with interconnecting communication capabilities to collaborate on solving a specific problem.

Supercomputers are applied to data-intensive and computation-heavy applications, often for scientific research or engineering calculations, including the following:

  • quantum mechanics
  • weather forecasting
  • oil and gas exploration
  • molecular modeling
  • physical simulations
  • aerodynamics
  • nuclear fusion research
  • cryptoanalysis

Supercomputer performance is measured in floating-point operations per second (FLOPS), not million instructions per second (MIPS). Supercomputers can contain tens of thousands of processors and perform billions or trillions of calculations per second. The most powerful supercomputers can perform up to a hundred quadrillion FLOPS. Also, as information is transferred between processors faster than distributed systems, supercomputers are ideal for real-time applications.

With reductions in cost during the 1990s, supercomputers began to find wider use across businesses for market research and other related modeling. In recent years, many organizations requiring large processing power have begun using cloud computing over supercomputers. High-performance computing (HPC) via the cloud is more affordable, scalable, and easier to upgrade compared with on-premises supercomputers.

History
1940s

Early computers, such as the Colossus and ENIAC, are sometimes referred to as supercomputers. However, the phrase "supercomputer" didn't come into use until the early 1960s.

Colossus, the first functional, electronic, digital computer designed by Tommy Flowers and built at Bletchly Park, began operations in 1944. The machine was designed to help British code-breakers during World War II working on a German cipher code-named "Tunny." The existence of Colossus was kept secret until 1975.

ENIAC, formally known as the Electronic Numerical Integrator and Computer, was the first vacuum tube, all-electronic, programmable computer to be unveiled to the world. Designed by John Mauchly and J. Presper Eckert in Philadelphia, ENIAC was revealed on the front page of The New York Times on February 14th, 1946. ENIAC was assembled and operated at the University of Pennsylvania. The machine was used to calculate complex trajectories for missiles and other armaments.

Announcement of ENIAC in the New York Times.
1950s

The Naval Ordnance Research Calculator (NORC) designed by IBM went into operation in 1954, becoming the most powerful computer in the world. Build between 1950 and 1954 at Columbia University's Watson Scientific Computing Laboratory, the machine helped forecast weather and performed other complex calculations.

In the late 1950s, the US government began regularly funding the development of cutting-edge, high-performance computer technology for military applications. These events would go on to have a significant influence on the progress of future supercomputers.

In 1951, supercomputer pioneer Seymour Cray joined Engineering Research Associates (ERA), which was later taken over by Remington Rand, Inc. (later merged with other companies to become Unisys Corporation). Cray left with William Norris (ERA’s founder), to begin Control Data Corporation (CDC) in 1957. At that time, Remington Rand’s UNIVAC line of computers and IBM had most of the market for business computers. Rather than compete for business computers, CDC chose to focus on fast, scientific computers.

On October 16th, 1959, CDC released its first model, designed by Seymour Cray—the CDC 1604. One of the first computers to replace vacuum tubes with transistors the CDC 1604 was quite popular in scientific laboratories and found applications in real-time data processing, controlling weapons systems, solving large scientific problems, and commercial applications.

1960s

During the early 1960s, a number of new computers specifically designed to be more powerful than the fastest commercial machines became available, and the phrase "supercomputer" came into use.

IBM

In 1961, IBM released its own scientific computer, the IBM 7030, commonly known as Stretch. The machine was capable of 2 million operations per second, but found few purchasers due to its tube and transistor hybrid design, and IBM temporarily withdrew from the supercomputer market after losses of $20 million.

CDC
Seymour Cray in 1963.

In 1964, CDC released the CDC 6600, which became the fastest computer in the world, capable of three million FLOPS. Often considered the first successful commercial supercomputer, CDC would go on to lead the commercial supercomputer industry.

The CDC 6600 was designed by Cray, who was deeply involved in every aspect of creating the computers that his companies built. In particular, he worked on the dense packaging of electronic components that computers require. His designs cut the distances signals had to travel to speed up machines.

UNIVAC

The UNIVAC LARC (Livermore Advanced Research Computer) was Remington Rand's first supercomputer. It was delivered to the Lawrence Livermore National Laboratory (then called the University of California Radiation Laboratory) in March of 1960.

Atlas

The Atlas computer was developed by Tom Kilburn at the University of Manchester, starting in 1956. The machine, completed in 1962, was the first computer to use virtual memory, had the first recognizable modern operating system, impressive RAM (96 KB), and multiprogramming capabilities.

Atlas computer at the University of Manchester.
1970s

Cray worked on a series of new supercomputers at CDC until he left to start Cray Research, Inc., in 1972. CDC continued producing supercomputers based on his designs. In 1976, Cray Research introduced the Cray-1 machine with the first successful implementation of vector processing.

On June 30th, 1976, Cray Research filed US patent No. 4,128,880 titled "Computer vector register processing," listing Seymour Cray as the sole inventor. The patent was granted on December 5th, 1978. The patent claimed a vector processing apparatus for a computer having a main memory and addressed shortcomings in conventional vector processing machines to enable efficient processing of short, moderate, and long length vectors. For this patent, Cray was posthumously inducted into the National Inventors Hall of Fame in 1997.

1980s

Introduced in 1982, the Cray X-MP was of the first machines to use multiprocessing linking two Cray-1 in parallel to triple their individual performance. In 1985, the Cray-2 a four-processor computer, became the first machine capable of over one billion FLOPS. In 1989, Cray left to form a new company, Cray Computer Corporation.

1990s

Announced in April 1989, NEC Corporation's SX-3/44R became the fastest supercomputer in the world in 1990. Based on NEC's SX-2 supercomputer from 1983, the SX-3 consists of seven models ranging from 1.4 GFLOPS uniprocessor to maximum performance of 22 GFLOPS quad processor model.

In 1993 the TOP500 project was launched to improve and renew the Mannheim supercomputer statistics. The organization ranks systems using the Linpack benchmark—the ability to solve a set of linear equations using a dense random matrix. Since 1993, TOP500 has published two lists a year (June and November) detailing the fastest supercomputers in the world. A timeline from TOP500 can be found here.

In March 1996, the Hitachi SR2201, a 1,024 processor, achieved 232.4 GFLOPS to become the fastest computer in the world. This was surpassed in June 1997 by the Intel ASCI Red supercomputer, the first teraflop/s computer. During the 90s when vector computers became less important, the US Department of Energy's ASCI (Accelerated Strategic Computing Initiative) program opened new funds with a focus on defense applications. Intel's ASCI Red used the cluster model.

On May 11th, 1997, IBM's Deep Blue became the first computer to beat a reigning world chess champion. Through six games, IBM Deep Blue won two games, lost one, and drew three.

21st century

Most modern supercomputers use the cluster model and many incorporate both CPUs and GPUs to work together to perform calculations. The following are the top supercomputer players, in terms of power:

  • Supercomputer Fugaku (Fujitsu)
  • Summit (IBM)
  • Sierra (IBM/INVIDIA)
  • Sunway TaihuLight
  • Perlmutter (HPE)
  • Selene (NVIDIA)

Supercomputer Fugaku became the fastest computer in the world in June 2020, with 442 Pflop/s exceeding the performance of Summit. The system is installed at Riken Center for Computational Science (R-CCS) in Kobe, Japan. Supercomputer Fugaku was co-developed by Riken and Fujitsu and is based on Fujitsu’s custom ARM A64FX processor.

The diagram below shows a timeline of key IBM supercomputers dating back to the Naval Ordnance Research Calculator (NORC) developed in 1954.

Timeline showing IBM's key supercomputer milestones.

Timeline

May 11, 1997
IBM's Deep Blue becomes the first computer to beat a reigning world chess champion.

Through six games, IBM Deep Blue won two games, lost one, and drew three.

1993
The TOP500 project is launched to improve and renew the Mannheim supercomputer statistics.

The organization ranks systems using the Linpack benchmark—the ability to solve a set of linear equations using a dense random matrix. Since 1993, TOP500 has published two lists a year (June and November) detailing the fastest supercomputers in the world.

1990
The SX3-4/44R, developed by the NEC Corporation, becomes the fastest supercomputer in the world.
1989
Seymour Cray leaves Cray Research Inc. to found Cray Computer Corporation.
1985
The Cray-2 computer becomes the first machine to exceed one billion FLOPS.
1982
The Cray X-MP is introduced, the first machine to use multiprocessing.

The machine connects two Cray-1 machines in parallel to triple performance.

December 5, 1978
Seymour Cray is granted US patent 4,128,880 titled "Computer vector register processing."

The patent claimed a vector processing apparatus for a computer having a main memory. For this patent, Cray was posthumously inducted into the National Inventors Hall of Fame in 1997.

1976
Cray Research introduces the Cray-1 machine with the first successful implementation of vector processing.
1972
Seymour Cray leaves CDC to found Cray Research, Inc.
1964
CDC releases the CDC 6600. The machine became the fastest computer in the world, capable of three million FLOPS.

The CDC 6600 is considered the first successful commercial supercomputer.

Further Resources

Title
Author
Link
Type
Date

A Brief History of Supercomputers

Jeff Layton

Web

Seymour Cray -- Father of the Supercomputing Industry

Web

November 23, 2009

Seymour Cray and 20th Century Super Computers

Web

September 10, 2020

Timeline of the MUSE/Atlas project at Manchester University,

1955 – 1971.

Simon Lavington

Web

November 2012

US4128880A - Computer vector register processing - Google Patents

Seymour R. Cray, Jr.

Patent

June 30, 1976

News

Search on Google
Search on Bing

References

Golden logo
By using this site, you agree to our Terms & Conditions.