Development of Computing Technologies
Danila Medvedev
Lappeenranta University of Technology
Information Technology Department
Seminar on Data Processing (Technical English)
Lappeenranta, 2002
The invention of the digital computer transformed our civilisation during the past 60 years and had profound social, economic and scientific effects. Still humans are just beginning to use the potential of computers. As our future progress will strongly depend on the development of computing technologies, it is important to clearly see the their possibilities.
The development of computing technologies depends primarily on the general technological level, available production methods and the market demand for faster computing. In the past there have been many recurring patterns in the development of computing technologies.
The aim of this paper is to outline the past development of computing technologies and specify the most important factors that influenced it. I describe the most important technological steps that made the evolution of digital computers possible. I explain what factors in the society and in the economy influenced the development of computing technologies.
Key Words: computing technologies, production, history, computer, microprocessor, integrated circuit
The computing technologies today are a very important part of our civilisation. The transistor (main component of today's computers) is often named the most important invention of the XX century and the computer is often named the most important invention ever. In 1982 its importance was acknowledged by the Time magazine, which selected the computer as its "Man of the Year".
As Michael Rothschild puts it: "[The] enormous information processing power [of the digital computer] has launched humanity into a new epoch of scientific discovery, economic evolution, and political interdependence. We have not experienced such a sea change in human affairs for 500 years, since Gutenberg's printing press set off an information explosion that led to the Scientific Revolution and, later, the Industrial Revolution." (12)
And still, the potential of computers is barely utilised yet. One of the important future advances that are foreseen today is general artificial intelligence. It will undoubtedly bring great benefits for the humankind. And it is sometimes argued that in less than a century the entire human civilisation will become digital.
It is therefore very important to understand clearly the prospects and possibilities of computing technologies. However, existing research does not provide this understanding. Numerous forecasts exist, but they rarely have solid and consistent structure, as I explained earlier (10). Many items in these forecasts are extremely vague. The items are frequently unconnected and have no effect or influence on each other. Even in the best forecasts (20) this problem is present.
The majority of mainstream professional futurologists also make another mistake. They often concentrate on making a variety of opposing scenarios, without specifying what factors are at play now, what are the existing trends and what will make them continue or reverse.
Professional "technologists"[1], as opposed to futurologists, usually concentrate on the possible (and likely from their professional point of view) outcomes, but also fail to take into account the present trends and factors influencing development.
As I suggested earlier (10), the problem lies in the absence of common framework for future forecasting.
The typical approach is well summarised by a quote from a "Futurology" radio programme in the IBM Solutions series (25):
There are any number of possible paths which the human race may take into the future and any attempt at forecasting that future is subject to the vagaries of unforeseen developments, techno-bottlenecks and changing social situations. Predictions which span the next five years or so are likely to be fairly accurate but beyond that the future is open to any intelligent guess. But until it arrives a little imagination can make it all the more interesting!
I strongly disagree with this. Although existing research confirms that for most specialists the forecast horizon is only 5-7 years (41), forecasting can and should be carried out scientifically and on a much longer timescale.
I wanted to create a framework for combining historical research, analysis of the present computer market and current technological situation and building the basis for scientific approach to forecasting of the future in the field of computing technologies.
In this paper I concentrate on creating a structured summary of the history of computer development and outlining the main factors that influenced it. The emphasis is done on computing hardware. Less attention is paid to developments in software, computer memory and peripheral devices.
First, I will outline the main events in the history of computing. Second, I will describe the main steps of technological development, which enabled the progress in computing technology. Third, I will analyse the main factors influencing the development of computers, concentrating on the demand for faster computers from various groups.
There is no clear agreement on the most appropriate term for the field that I am describing. "Data processing" and "information processing" make a strong emphasis on the data and information respectively (26). "Computational technologies" can only be used to describe technologies that use computation. Finally, there is a term "computing technologies". It can have two meanings:
1) Various methods of using computer for calculations in various scientific areas, such as biotech, astrophysics, etc.
2) The technologies of calculating and building machines for calculating.
In this paper I used the term "computing technologies" in the second meaning.
The moment of beginning of computing is hard to specify precisely. Some authors place it as early as 350 mln years ago, when the first tetrapod evolved the first fingers (8). Fingers were later used for counting, which is reflected in the fact that the word "digit" is applied both to fingers and to numbers. Others say that computing started with the development of numbers or with early shamanism practices (5).
But the real computing began with the development of the abacus. It allowed users to perform complex arithmetical operations by keeping track of the current state of calculation.
The abacus remained state of the art computing device until 1612, when the logarithm was invented by a Scottish mathematician John Napier. He also created a calculator, called Napier's Bones, consisting of several numbered sticks, that were put next to each other and moved to perform division and multiplication. Ten years later a modification of this system, known as slide rule, was created by William Oughtred. Slide rule remained in use by scientists and engineers for the next 350 years (8).
Next important invention was the Pascaline, a mechanical multiplier based on gears, designed by Blaise Pascal (1623-1662). The machine was capable only of addition, but simple techniques, allowed performing other operations, such as using complimentary number for subtracting. In 1642 Pascal made another version of calculator that used falling weights instead of gears (4).
Thirty years later a German mathematician Gottfried Wilhelm von Leibniz (1646-1716) designed an upgrade for the Pascaline. Leibniz wheel was a multiplication-capable machine, based on a stepped drum mechanism (4).
Still, it wasn't until XIX century that arithmetic machines became widely used. In 1820 Thomas de Colmar designed a first commercial calculator named Arithmometer, based on Leibniz wheel (4).
In early XIX century English mathematics professor Charles Babbage (1791-1871) noticed the need to check large amounts of calculations for the Royal Astronomical Society. Babbage saw the repetition of the work and connected this with the ability of machines to perform repetitive tasks. He began work on a machine that would be capable to solve differential equations. The design included about 25 000 parts. The machine, as well as the second version and Analytical Engine used gears for calculations. Total weight of the machine would be 15 tonnes and the size would be 2.4 x 2 x 1 meters (height, length, depth). For design and construction of the machine Babbage got about £17,500 from the government. First design work started in 1820. In 1832 first part of the machine was built but work on the project was halted in 1833 (34).
After ten years of working on this project, he came to the idea of a more versatile machine. With the help of Augusta Ada King, Countess of Lovelace (1815-1842) he started working on a first ever general-purpose computer, which he called the Analytical Engine. It included conditional control (similar to IF operator) that allowed it to execute non-linear programs. The machine was designed to be completely mechanical and to be powered by steam engine. The machine consisted of more than 50 000 parts. The machine accepted inputs on punch cards. It had internal memory of 1,000 numbers of up to 50 decimal digits long (approximately 20Kb) (22). The cycle time was approximately 3 seconds per operation (4). This machine was never built.
Babbage designed a new version of Difference Engine between 1847 and 1849, utilising the experience from working on the Analytical Engine. This allowed him to create a simplier design that required approximately three times less parts. Babbage didn't complete this machine, but a working model was built in 1991. In 1993 the Science Museum tested the Difference Engine 2 against a modern computer. Canon BJ Notebook BN22 (with a TI486slc 25MHz processor, running custom-made Windows software) was tested against Difference Engine 2 for calculating a seventh order polynomial to 31 figures of precision. It is difficult to compare the speeds of two machines exactly, but the Canon computer was 5-50 times faster (36).
Another important invention, although not directly related to computing, happened in the end of XIX century. For the 1890 U.S. census new data processing machines were created by Herman Hollerith (1860-1929). The tabulating machines used punch cards to read census data and process it. While not directly related to computation, this was an important step in the development of data-processing machines. The tabulators became very successful and Hollerith later formed a new company to manufacture and sell them. This company was later renamed to IBM (International Business Machines) and became one of the largest players on the computer market (5).
In 1935 Vannevar Bush (1890-1974), an American scientist, developed Differential Analyzer, a large analogue electric calculator capable of solving differential equations. It used 2 000 vacuum tubes, 150 motors and weighed 100 tons.
In 1936, a British mathematician Alan Turing in en effort to describe all mathematical problems that are logically possible to solve (5) described the Turing Machine – a hypothetical programmable computer. He designed it to test one of the Hilbert problems – the Decision Problem (Entscheidungsproblem). The Turing Machine consists of an infinitely long tape, a head that can read and write to this tape (1s and 0s) and a program, that is usually represented by a state graph that uses a symbol on the tape under the head as input and outputs by writing a symbol to the tape and moving the head in either direction. Every computer built or designed to date is equivalent to some Turing Machine (14).
Turing also described a Universal Turing Machine (UTM). This is a Turing Machine, which is capable of emulating any other Turing Machine, given a finite length tape with a program (6). The concept of UTM is extremely important for today's computing and also for modern and future advanced developments, such as quantum computers, possibility of hypercomputers, question of the computational complexity of the Universe and possibility of modelling of human mind on a computer.
First electronic computer, ABC (Atanasoff-Berry Computer) was built in 1940 by John Vincent Atanasoff (1903-1995), a professor at Iowa State College and Clifford Berry, his graduate student. This was the first binary computer that combined Boolean algebra with on/off states of electronic circuits (15). However, ABC was not a general-purpose computer. It was hardwired for solving systems of linear equations (19).
Unfortunately, the ABC didn't have much effect on the field of computing at that time. Other machines were still electromechanical calculators using electromagnetic relays to control mechanical parts, such as counter wheels. In 1943 British engineers developed a specialised code-breaking machine named Colossus. In 1944 Americans created Mark I, technically similar machine for calculating ballistic charts. These machines were very large (Mark I was half as long as a football field) and slow (it took Mark I about 3-5 seconds per calculation) (22).
The race to build a first computer began in the middle of 1930s. In 1941 Konrad Zuse (1910-1995), a German engineer, created the first programmable computer named Z3 (several less successful versions were constructed earlier). It used a binary system for data representation, was built completely out of relays (600 for the arithmetic unit and 1,800 for the memory and control units) without mechanical parts and was controlled by perforated strips of film. Z3 was the first general-purpose electronic computer (17).
ENIAC (Electrical Numerical Integrator and Computer) was a large digital electronic computer built in the United States in 1946 by John W. Mauchly and J. Presper Eckert. It used not binary, but decimal system. It contained about 18 000 vacuum tubes and took up more than 170 sq. m. Previous computers never used more than 2000 vacuum tubes. The programming was done by rewiring a dedicated part of the computer. Contrary to popular belief, ENIAC was not the first electronic computer, as it was built five years after Z3.
In 1947 John von Neumann completed the design of von Neumann architecture, which is used in most modern computers. The most important feature of this architecture is that the programme code is data and can be stored in the same memory that holds variables and other data.
EDVAC (Electronic Discrete Variable Automatic Computer) was designed and built in 1949-1951 by Mauchly and Eckert, the creators of ENIAC. The main difference between these two machines was the ability of EDVAC to store programs in electronic form. This was made possible by the large amount of internal memory in EDVAC, more than in any other computing device at that moment (5). Later, in 1952 von Neumann developed MANIAC (Mathematical Analyzer, Numerical Integrator and Computer), which was similar to EDVAC, but faster. By 1953 there was already about 100 computers in the world (16).
In 1956, nine years after transistor was invented, a second generation of computers became possible. Among the first machines based on transistors instead of vacuum tubes were two supercomputers, Stretch by IBM and LARC by Sperry-Rand. They were followed by a number of other models. One of the most commercially successful was IBM 1401. By 1965 second-generation computers were used by many large companies (22).
3rd generation of computers was based on the integrated circuit (computer chip), developed in 1958. First computers of the new generation became available in 1965.
In addition to computers, integrated circuits quickly became used in various devices, including audio and video equipment, automobiles, home electronic equipment, white goods, communication devices, industrial electronics and others.
In 1964 IBM introduced its System/360 family of computers based on a proprietary technology similar to integrated circuits. These machines became a very important step in the computing history and introduced a number of new concepts. All System/360 computers used the same assembler language. For the first time programs written for one model could be run on another without modifictions. The concept of upgrade was introduced that allowed companies to replace a computer with a more powerful one, while retaining old software. The System/360 family became a huge commercial success for IBM.
4th generation of computers, based on a microprocessor, appeared in 1972, just one year after the first microprocessor was created. In 1972 Micro Instrumentation and Telemetry Systems (MITS) released the first digital microcomputer, MITS 816, based on Intel 8008 microprocessor. Two years later, several other companies introduced their microcomputers, including Scelbi and Mark-8. But the microcomputers market really took off in 1975, when MITS Altair 8800 was released. This computer, sold for less than 400 dollars, included Intel 8080 CPU and 256 bytes of RAM. The machine code programming was done by flipping switches on the front panel.
In 1977 Steve Jobs and Steve Wozniakintroduced Apple II, that had colour graphics, built-in BASIC and 4Kb memory. The price was 1300$ (5). In the same year 1977, Tandy Radio Shack introduced TRS-80, with 16Kb RAM, Z-80 processor and monochrome display. The later model had 64Kb RAM and a floppy disk drive to store programs and data (1).
IBM released its first personal computer, called IBM PC, in 1979. It had 16Kb of RAM and connection to tape recorder. It was one of the first computers to use Intel x86 architecture. The price of the first IBM model was 1265$.
The IBM PC standard was widely adopted by consumers because of IBM strong position on the market of business computers. Another important factor was the relatively open modular architecture. Later the IBM PC BIOS was reverse engineered and production of "PC clones" became possible and even encouraged by IBM. This drove the prices down and made IBM PC compatible machines the standard.
The PC specification from IBM required that all IBM PC compatible machines use Intel or Intel-compatible processor. This guaranteed the compatibility on the machine code level for all IBM PC computers (33).
In 1980s personal computers became very popular. The number of personal computers skyrocketed. In 1981 there were 2 million PC in the world and more than 5.5 million in 1982, just one year after. Ten years later, in 1992 65 million PCs were being used world-wide (22). Ten more years into the future, in 2002 there are about 287 million of personal computers used (3).
In addition to computers, microprocessors became increasingly used in various electronic devices, often replacing custom-designed chips.
But microprocessors were also used to power world's most advanced computers. In 1976 Cray-1, an 8.8 million dollar supercomputer was installed by Cray Research in Los Alamos National Laboratory. Cray-1 had the speed of 160 megaflops and 8Mb main memory. Cray Research remained the leading producer of supercomputers, with its 1988 model Cray Y-MP reaching 1 gigaflop and Cray T3E-1200E hitting 1 teraflop in 1998.
On the other end of the spectrum, in 1993 Apple released one of the first personal digital assistants (PDA) under the name Apple MessagePad (also known as Newton). A small (2.8 x 11.8 x 20x8) handheld device featured a 160MHz processor and 4-8Mb of RAM. But PDAs remained a niche product, until Palm Computing released cheaper and smaller device in 1996 under the name PalmPilot. After that handheld computers took off and in 2001 more than 13 million devices were sold (9).
Now handheld computers are another important platform where microprocessors are used. The key requirements for the CPUs is low power consumption that allows to extend the battery life of the device and saves power for other features, such as hi-res color displays and communication features. In addition to PDAs, microprocessors are now used in mobile phones, MP3 players and other electronic devices.
During 1971–2002 Intel was constantly releasing faster and more complex microprocessors, from 4-bit 108KHz 4004 up to 32-bit 2.4GHz Pentium IV and 64-bit Itanium processors. During this period Intel encountered some competition from several companies, including Cyrix, Transmeta, Via and AMD (Advanced Micro Devices), the strongest Intel's competitor today.
In the 1990s the motivation to upgrade to faster computers was often provided by Microsoft Windows operating system. This family of bloated and inefficient products introduced a practice of forced upgrade when every next version required a next generation computer.
Soon another reason for frequent upgrades appeared. In 1993 id Software released Doom – one of the most successful PC games in history. Before Doom, games were usually designed to run on most existing computers, regardless of their speed. But advanced graphics of Doom required a fast processor to maintain sufficient framerate for quality gameplay. The ability to run Doom well was an important factor in buying a new personal computer at that time.
The power of CPU was not sufficient for the rapidly evolving 3D games. In 1996 3dfx released Voodoo – the first 3D acceleration video chip for consumer products. One of the most popular cards based on this chip was Diamond Monster 3D from Diamond Multimedia. The 3D accelerators, now known as Graphical Processing Units (GPU), greatly improved the performance of many popular games, bringing to the consumer PCs the quality of graphics available before only on expensive high-end graphics workstations.
Another potentially unlimited market for processors is consumer robotics. The first successful consumer robotic product was launched by Sony in 1999. AIBO (ERS-110) robotic pet was the first advanced entertainment robot. During 1999-2002 Sony sold hundreds of thousands of AIBOs at an average price of 2000$. ERS-110 used a 100MHz RISC processor and had 16Mb of internal memory.
After more than 60 years of digital computer history, the total number of computers in the world is estimated at 287 million. According to Jack Campbell (3), they included:
The single most powerful computer, more powerful than other 19 computers from the TOP20 was completed in 2002. NEC Earth-Simulator, the most powerful supercomputer in the world was put into operation in Japanese climate research centre. The speed of this supercomputer exceeds 35 teraflops (39), which, according to estimates by Hans Moravec (11), is approximately 3 times greater than the computational complexity of human brain.
As we see from the previous section, the technology used in computing devices underwent a significant evolution since first of them were introduced.
The idea of computing originated very early. First computing devices were purely mechanical, because that was the available technology at those times. The details of these calculators required high precision. The only was to achieve it was by using qualified manual labour. This made such calculators too expensive to become popular.
By XIX century, new production methods finally allowed the production of affordable mechanical calculators. But since all these devices were operated manually, the next goal became to eliminate the potential for human error. Human actions could not be completely accurate and fast enough. So, the first important step was to make the devices "self-functioning", so that they required only some power source (steam engine or manually rotating some handle) and information (data and programs) input. The program was input either by configuring the mechanical parts prior to launching the device or during the runtime, but while the computing device was running, it didn't require further assistance. That was done by having internal memory, where intermediate results were stored.
But the technology of the XIX century had its limits. Manufacturing parts for the automatic machines designed by Babbage "stretched the standards of engineering practice of the time… [The] mechanisms demanded hundreds of near-identical precision parts" (34). But at that time there was no way to automatically manufacture them.
The limitations in the machine tool technology of that time are most often provided as a reason for Babbage's failure. But in 1991 the Science Museum constructed a working Difference Engine 2, according to original design drawings. The details were manufactured with precision similar to what was achievable in the middle of XIX century. This machine has 4 000 moving parts and weights 2.6 tonnes (35).
But Michael Rothschild suggests that Babbage failed to build his computers, because the decimal number system that he used required too great precision, as opposed to binary system, used in most successful computers (12).
By the beginning of the XX century numerous inventions and improved production processes combined together made the creation of the mechanical computers possible. But the mechanical devices were bulky and slow. For example, the Differential Analyzer, an electromechanical calculating machine made in 1935 weighted more than 100 tons. The size and speed were the main reasons why everyone looked into other possibilities.
The transition from mechanical logical circuits to digital computing was gradual. The intermediate step was the electromechanical telephone relays, such as used in German Z3. The Z3 used a binary system, and was built using Boolean algebra (devised by George Boole in 1854) that existed for many years and was well understood by that time. Z3 was constructed completely out of relays, without any mechanical parts. Programs were fed into Z3 on perforated film strips.
A few years before Z3 John Atanasoff realised that to build a binary computer vacuum tubes with their on and off states can be used very effectively. Vacuum tubes, invented in 1906 by American physicist Lee De Forest (1873-1961) (5), were faster and smaller than mechanical (and electromechanical) parts. First generation of computers used vacuum tubes for building circuits and magnetic drums for storage.
All new computers also used binary system with the exception of the ENIAC, as binary numbers simplified the design of the arithmetic units.
But vacuum tubes were prone to failure, required a lot of power, efficient cooling and were taking a lot of space. Another alternative was clearly needed.
In 1940 research was carried on improving the radar technology. Radars at that time used semiconductor crystal for translation of reflected (received) radio waves into direct current that allowed visualisation of radar signal on the screen. Research to improve the qualities of the semiconductor crystals was done by many institutions, including Bell Labs. After the World War 2 increased knowledge of solid state physics hinted Bell administration that semiconductors can be used to make a better amplifier to replace vacuum tubes in Bell's phone networks (40).
In 1947 three scientists from Bell Laboratories, William Shockley, John Bardeen, and Walter Brattain, invented the transistor. Transistors required much less space and electrical power and quickly replaced vacuum tubes in computing devices, but as the complexity of computers increased, it became increasingly difficult to design them by manually wiring all the transistors. In addition to that, transistors could no longer be reduced in size, since it was still necessary to manually connect them by wires.
In 1958, Jack St. Clair Kilby from Texas Instruments simultaneously with Robert Noyce from Fairchild Semiconductor invented the integrated circuit (also called a computer chip). An integrated circuit is a prefabricated silicon structure, containing interconnected transistors and other elements such as capacitors and resistors. This permitted the mass production of parts for computing devices, increased the processing speed and decreased transistor size (5). The original integrated circuit was only a few centimetres in size and contained one transistor, three resistors and one capacitor (2). Both companies were able to patent the integrated circuit and after a few years they made a cross-licensing agreement.
At that time noone realised all implications of the integrated circuit. Jack Kilby later wrote: "What we didn't realize then was that the integrated circuit would reduce the cost of electronic functions by a factor of a million to one, nothing had ever done that for anything before" (2).
System/360, introduced by IBM in 1964 was not based on the integrated circuits. As IBM was unsure about the new technology, they used a hybrid approach called Solid Logic Technology (SLT). In SLT only passive circuits were fabricated on the ceramic substrate. Semiconductor elements (diodes and transistors) were manufactured separately using standard second generation production techniques (23).
There was little market pressure to improve the situation. But when Intel (former Fairchild Semiconductor) got an order from a Japanese client that required designing about 12 separate chips, Intel engineers decided that it might be easier to make one big programmable general-purpose logic chip that would integrate several functions (central processing unit, memory, input and output controls).
In 1971 Intel released the first microprocessor, developed by Marcian E. Hoff. The device 0.3 x 0.4 mm in size contained 2 300 transistors and executed 60,000 instructions per second, as many as ENIAC, that had filled 300 cubic metres with 18 000 vacuum tubes.
It was a clever shortcut. Replicating the whole computer on a chip at that time would require at least 20 000 transistors. That was not possible with the available technology. Several companies designed complex solutions consisting of several chips, but none was successful. Intel 4004 was both simple enough to be manufactured, but proved to be powerful enough to be used in many devices. That permitted gradual development and soon microprocessor outcompeted traditional third-generation computers.
The development of microcomputers followed a pattern similar to that of the microprocessor. While initially the capabilities of the microcomputers were minimal, compared with the mainframes, the microcomputers developed at much faster pace and soon made the mainframes obsolete.
In both cases a potentially superior technology could not compety directly with a mature technology. Instead a large side-market was needed to support the rapid development that allowed new technology to win the competition.
Since the 1940s the average time it required for a new invention to find its way into successful commercial product was constantly decreasing. The transistor was invented in 1947 and in 1956 first second-generation computers that used transistors were built. In 1958 the integrated circuit was invented and in 1964 first third-generation computers were built. Finally, in 1971 the microprocessor was invented and in 1972 first fourth-generation computer (microcomputer) was built.
4th generation became the last. No other technology had replaced the microprocessor. But the microprocessor proved to have an enormous potential. In 1965 Gordon Moore, the founder of Intel made his famous observation, which is know widely known as "Moore's Law". Originally Gordon Moore observed the exponential growth in the number of transistors placed on a chip and predicted that this trend will continue. In 1970s the number of transistors on a chip was doubling every 2-3 years. In effect, Moore's Law means that the performance of the microprocessors grows exponentially. This Law still remains true in 2002, 37 years after it was first observed (30).
In 2001 Ray Kurzweil made two important observations regarding the Moore's Law. First, he notes that Moore's Law is the fifth computing paradigm providing exponentially increasing price/performance, first four being mechanical devices, relays, vacuum tubes and transistors. His second observation is that when plotted on a logarithmic scale graph, the price/performance function takes form of another exponential curve, making the growth in computing price/performance double exponential (7).
Many technologies that can potentially become the new paradigm were under development. In 1988 the first optical chip was developed, that used light instead of electricity (16). The developments in optical computing can allow a dramatic increase in computer performance. In 1994 the first demo of DNA computers was done by Leonard Adleman. The DNA computer was used to solve the travelling salesman problem. The speed of computations that inherently parallel DNA computer is capable of is enormous. During the 1994 demo calculations were carried out with the speed of 100 teraflops, 3 times faster than Earth Simulator, the most powerful supercomputer today. In 2002 Olympus Optical Co. "…announced the successful development of the world's first functional DNA computer…" (32). No details about its capabilities have been given yet.
Another promising technology is quantum computing. In 1998 the first 2-qubit quantum computer was demonstrated in University of California. In 2001 a 7-qubit computer was created (27). Since the computing power of quantum computers grows exponentially with their size, a 30-qubit quantum computer would be as powerful as the fastest supercomputer today (29).
But future technologies might not be used in CPUs. Other applications became increasingly computationally intensive. After the introduction of Voodoo chip the development of the 3D accelerator technology continued at a very past pace and in modern computer the GPU is more powerful than the CPU. In 2002 GeForce4, the latest graphics chip from nVidia, has 63 million transistors, compared with 37.5 million in AMD latest Athlon XP and 55 million in Intel's new Northwood Pentium 4 (31).
For many centuries merchants and bankers used abacus for all their calculating needs. Only after the cheap mechanical calculators became possible in XIX century, they switched to more advanced devices. But since the mechanical technology was bulky and relatively fragile, abacus wasn't completely abandoned until the middle of the XX century (and as late as 1990s in some countries), when the digital electronic calculator was invented.
By the end of XVIII century science in most European countries was already institutionalised to some extent. Understanding of long-term scientific goals created demand for calculation tools, first of all, to provide accurate logarithmic tables. But the process of calculating these tables was extremely time-consuming and prone to mistakes. There were many versions of logarithm tables available that often had quite different values. The process of checking and recalculating the tables could very well take a whole life of a mathematician.
Other than for recalculation of logarithms tables, in early XIX century there was no demand for computers. Science didn't knew the natural laws required for simulations yet, military did not have accurate artillery that would require complex calculations, businesses were so small that even arithmometers were sometimes too powerful.
In the end of the XIX century for the first time organisations such as U.S. Census Bureau needed to process large amounts of data. Although, strictly speaking, this demand was not for calculations, it was quite important for the development of computing technologies.
In 1930s a lot of research was carried out for the military, especially for automatic calculation of artillery firing tables. Many custom mechanical devices were designed and created. Although the general technology to build a general-purpose mechanical computer was available, large scale attempts to construct the computer began only with the onset of the World War 2.
The prospect of war created demand for intensive calculations in various areas. Powerful calculating devices suddenly became needed for design of aeroplane wing, for breaking enemy secret codes and for calculating ballistic tables for the artillery.
While a skilled person could spend as much as 20 hours for calculating a 60 second trajectory using a desk calculator, the differential analyser could produce the result in less than 15 minutes, while a fast general-purpose computer such as ENIAC could finish the job in 30 seconds (12). There was clearly large unmet demand for automated computing.
"During the period 1948-1955, when it was retired, ENIAC was operated successfully for a total of 80,223 hours of operation. In addition to ballistics, fields of application included weather prediction, atomic energy calculations, cosmic ray studies, thermal ignition, random-number studies, wind tunnel design, and other scientific uses.
Significantly, the Army also made ENIAC available to universities free of charge, and a number of problems were run under this arrangement, including studies of compressible laminar boundary layer flow (Cambridge, 1946), zero-pressure properties of diatomic gases (Penn, 1946), and reflection and refraction of plane shock waves (IAS, 1947)." (12)
The first computer used in atomic research was MANIAC, which was used for calculations of Mike, the first hydrogen bomb. Two supercomputers – first machines based on the transistor technology, Stretch by IBM and LARC by Sperry-Rand were developed especially for atomic energy laboratories. The explosion of nuclear bomb was extremely computationally-intensive task, but in the end of the XX century it became possible to simulate it on supercomputers with sufficient details. In December 2002 five among seven fastest supercomputers are used in atomic research (39).
"Early on in the computers history, they were seen as devices having a strictly scientific role. This mindset helped relegate computers to Government and Military uses. … [The] reason for the lack of … commercial interest was the fact that investors were hesitant about offering large amounts of capital to projects that people were unsure of." (37)
System/360 was the first series of compatible computers. This provided great flexibility for the customers, as a computer could now be upgraded to a more powerful model retaining old data and software. IBM began marketing the family of System/360 as business tools (37).
"Although transistor technology could allow for the construction of smaller computers, many felt that instead of going small, larger and more powerful computers could be built" (37).
It was already realised that computers had a lot of potential. Altair was bringing this potential to the individual user.
"The number of computers in U.S. homes went from a few hundred in 1975 to over 20 000 just one year later. Still, the driving force behind the expansion remained in the hands of hobbyists who would emerge from their garages with new developments that propelled the industry and rendering new technology obsolete in just a few months.
One problem with the computer was its limited range of uses. It was the emergence of the word processor and the modem that created domestic uses aside from video games and mathematical calculations" (37).
In addition to good technology and good product a strong company to push them was needed. The personal computers became really popular only after the IBM introduced their architecture. Business users were attracted to the new platform by increased compatibility that allowed transferring data and software between different machines without recompiling the programs for each machine.
A similar example is Sony AIBO. To make the idea of an entertainment robot accepted, a backing of a large corporation was needed.
The personal computer was targeted at both companies and private users. First office applications, such as word-processing, electronic spreadsheets, databases (WordStar on first microcomputers, MacWrite, LisaCalc, VisiCalc on Apple, Microsoft Word, WordPerfect, Lotus 1-2-3 and dBase on PC) allowed to automate and accelerate a lot of office work.
Storing documents in electronic format, where they can quickly and easily be modified revolutionised document-editing. Electronic spreadsheets that allowed to automatically recalculate complex sets of formulas revolutionised business calculations.
Then the number of applications increased. The common Intel/PC/DOS/Windows architecture solved many problems of compatibility and created mass-market for the software. The attractiveness of the computer was continually increasing for business and private customers.
The development of handheld computer marked a new paradigm of computer use – constant interaction with computers. This idea later developed into a concept of pervasive computing.
Since Doom was released in 1993, games became a crucial part of the PC market. The games industry surpassed the movie industry in terms of annual sales. In United States more than 50% of the population play computer and video games. While some claim that consoles offer better price/performance ratio, the main difference is actually in the payment model (consoles are sold at a loss while the revenue is generated through the sale of the games). Special attention is paid to the attractiveness of the PC as a game platform, as games are very important to provide the incentive for upgrades.
Many credit id Software with accelerated development and acceptance of computer technologies and 3D graphics in particular. Like IBM with the PC, id Software was able to push customers to upgrade their computers. With Quake 3 id Software decided to drop the software rendering mode, essentially making 3D acceleration a standard function of any modern computer. Although today they no longer have a similar degree of control over users' hardware, it is expected that they will again push the "minimum requirements" for a PC with their upcoming title Doom 3.
With the growing speed of the CPU as well as memory and storage capacities, new unexpected applications became possible for the PC. First MP3 playback demanded a certain minimum speed from the processor. Later digital video formats such as DivX set new minimum requirements. In the same way, the growing popularity of the digital photo and recently video cameras make photo and video editing standard features of a modern PC, further increasing the minimum acceptable processor speed.
In the past personal computers were changing in a more revolutionary way. When a new generation of computers was becoming available, this offered really good reasons to upgrade.
A typical upgrade path could look like: PC AT/XT ® 286 ® 386/486 ® Pentium
Pentiums (with speeds around 200MHz) already got all attributes of a modern computer. They had CDs, sound cards and 3D accelerators. They could (and still can) ran Windows 95/Windows 98 operating system, which is still mostly compatible with the latest Microsoft offerings. The main changes after the Pentium were not in radically new features, but purely in speed.
Below I list some of the factors that force people to upgrade in such situation.
Because of these factors, today in the United States the average life of the PC is merely 2 years. Still, most users are comfortable with the idea of periodic upgrades.
Another important factor is that there doesn't seem to be sufficient demand for cheap slow CPUs. Since in budget-models the CPU costs less than 50$, noone really wants to save on it. And when upgrading something (memory, videocard, etc.) the computer is often changed for a new one or upgraded completely with the new CPU. When Cyrix was acquired in 1997 by National Semiconductor, new strategy concentrated on low-end PCs, including "supermarket PCs", "set-top boxes" and similar areas. This was apparently a bad business decision for Cyrix that soon was forced to stop its operations.
The latest generation in computers, the fourth generation, continues since 1971. In early 1980s 5th Generation Computer Systems Project, a joint Japanese-European venture started. The goal of the project was to develop a new generation of computers based on artificial intelligence. Unfortunately, no significant results were produced. The cheap and powerful personal computers of the 1990s became much more successful than any spin-offs from the 5th generation project.
In 1992 the Real World Computing Project (RWCP) or the 6th Generation Computing Project was launched, again as a collaboration between Japanese and European researchers. This time the project was much broader and less focused. The RWCP is related to soft computing methods, such as neural nets or fuzzy techniques (18).
AIBO robots paved the way for smart toys, a wide range of interactive products from different manufacturers. Today customers demand increased interactivity from many products, including toys. Next step after creating an interactive device is making a smart device that uses artificial intelligence techniques to adapt to the changing situation without direct user input.
The development in the area of supercomputers is based on the ever-increasing need for fast calculations. Many industries and research areas need powerful supercomputers for drug design, DNA sequencing, virtual car crash-tests, simulations of astronomical objects, protein molecules, weather and climate.
The digital computers were actively developing since the middle of the XX century. Their development started in response to emerging need for high-speed cheap and reliable computation. Through several paradigm changes the speed of computers was constantly increasing in order to provide sufficient computing power for a variety of different purposes.
Since the beginning of the 4th generation of computers a stable development model has developed. It incorporates a quasi-monopoly of Intel, the company that forces itself to maintain a steady rate of progress by creating and reinforcing certain market expectations.
The development of computing technologies strongly depends on the general technological level of the economy and on the available production methods.
A number of development patterns can be specified in relation to the computing technology. For example, in both cases of microprocessor and microcomputer a potentially superior technology aimed at a wide unoccupied market, developed itself and replaced an old mature technology.
Microprocessor (or any other general-purpose computing technology) can be used in a variety of different areas. New areas are always possible.
Initially the demand for computing has been low, but after it started growing in the early XX century, the computing technologies were not able to satisfy this demand.
Computing technologies were first seen as purely scientific related. Later possible usage areas included military and government tasks. After some time, the field of computing was expanded again to include business needs. Finally, first computers were created that targeted individual customers. This suggests that the broadest possible application should always be considered. Special attention must also always be paid to the specific needs of different groups.
There is a number of key application areas that can generate the greatest demand for computing power. Among them are:
And since computers are universal, completely new applications are always possible.
The volume of this paper placed some limitations on the research. In the future the development of software industry and software technologies can be analysed. Another area of interest is telecommunications.
An overview of today's computing industry (chip manufacturers, computer manufacturers, components manufacturers, GPU makers, software industry, R&D, supercomputers) should be made to test the conclusions of this paper and expand them. The factors that influenced computer development in our history can be applied to today's situation. An overview of the main computer users today is needed (industries, social categories).
Next direction is research of potential future developments. Existing forecasts should be analysed and tested when possible. The analysis of today's trends can be carried out to determine how the can influence the future development. Among the questions to consider are distribution of research roles, speed of development, analysis of dependencies in technologies. Last two questions can be combined to develop a future timeline (development map).
Scenarios can be developed, describing the use of processing power in the future. Existing long-term forecasts can be validated. It can be interesting to analyse philosophical consequences of computing technologies development, such as the nature of the human mind, conscience, possibility of AI, as well, as questions about the laws of nature and the structure of our universe.
In addition to qualitative analysis, quantitative data on computing technologies development can be collected. It should include speeds of various computers in history, using an objective measure of speed such as MIPS or FLOPS, price/performance ratios, total available processing power, number of computing devices. Using this quantitative data the relative importance of the main factors influencing the computing technologies development can be verified.
1.
anomie,
2000, TRS-80
Model I, Everything2
http://www.everything2.com/index.pl?node_id=523269
2.
Bellis
M., The
History of the Integrated Circuit, About.com
http://inventors.about.com/library/weekly/aa080498.htm
3.
Campbell
J., 2002, Shattering
the five percent myth, SpyMac
http://www.alliedcomputing.com/default.php?news_id=50
4.
Edwards
P., 1997, History of Computers
http://www.si.umich.edu/~pne/STS161/Misc_files/Lectures/19th_century.rtf
5.
Hoyle
M. A., 2002, The History of Computing Science
http://www.eingang.org/Lecture/toc.html
6.
JerboaKolinowski,
2002, Universal Turing Machine, Everything2
http://www.everything2.com/index.pl?node=Universal+Turing+Machine
7.
Kurzweil
R., 2001, The Law of Accelerating
Returns
http://www.kurzweilai.net/articles/art0134.html?printable=1
8.
Lake
M., 2000, A Brief History of Computing, Everything2
http://www.everything2.org/index.pl?node_id=816527
9.
McDonough
B., 2002, Global PDA Sales Grew 18
Percent in 2001, Wireless NewsFactor
http://www.wirelessnewsfactor.com/perl/story/16354.html
10. Medvedev D., 2002, Planning for the Future, International Management Symposium, St. Gallen, Switzerland, http://www.lut.fi/~medvedev/planning/future.html
11. Moravec H., 1988, Mind Children
12. Moye W. T., 1996, ENIAC: The Army-Sponsored Revolution, Army Research Laboratory
13.
Rothschild
M., 1992, Hail Zuse! Heed Zuse!, Upside
http://www.bionomics.org/text/resource/articles/ar_015.html
14. rp, irexe, 2001-2002, Turing Machine, Everything2
http://www.everything2.com/index.pl?node=Turing+Machine
15. tes, 2001, John Atanasoff, Everything2
http://www.everything2.org/index.pl?node_id=776867
16. White S., 2002, A Brief History of Computing
http://www.ox.compsoc.net/~swhite/history/timelines.html
17.
Zuse
H., The
Life and Work of Konrad Zuse
http://www.epemag.com/zuse/part4a.htm
18.
AI, Japan Inc.
http://www.japaninc.com/article.php?articleID=514
19. Atanasoff Berry Computer, Wikipedia, 2002
http://www.wikipedia.org/wiki/Atanasoff_Berry_Computer
20. BT Timeline, British Telecom
21.
Calculating with DNA, Bio IT World, 2002
http://www.bio-itworld.com/archive/100902/dna.html
22. Computers: History and Development, Jones Telecommunications and Multimedia Encyclopedia, http://www.digitalcentury.com/encyclo/update/comp_hd.html
23.
Computer Structures: Readings & Examples, Microsoft Research
http://research.microsoft.com/users/gbell/Computer_Structures__Readings_and_Examples/00000584.htm
24.
Cray Inc. History, Cray Inc.
http://www.cray.com/company/history.html
25.
"Futurology", IBM Solutions radio series, Science
Line, Somethin' Else
http://www.sciencenet.org.uk/soundofsci/future.html
26. Google Glossary
http://labs.google.com/glossary
27.
IBM's Test-Tube Quantum Computer Makes History, IBM Research News, 2001
http://www.research.ibm.com/resources/news/20011219_quantum.shtml
28.
Integrated Circuits, Transistorized!, 1999
http://www.pbs.org/transistor/background1/events/icinv.html
29.
Leaps in Quantum Computing, DC Physics, 2000
http://www.dctech.com/physics/features/0400.php
30.
Moore's Law, Intel
http://www.intel.com/research/silicon/mooreslaw.htm
31. nVidia GeForce4 Ti 4600 Review, ZDNet
http://reviews.zdnet.co.uk/review/14/1/158.html
32. Olympus Press Release, Olympus Optical, 2002
http://www.olympus.co.jp/Special/Olp/Info/n020128E.html
33. Personal Computer Software: Word Processors & Spreadsheets
http://library.utmem.edu/HSLBC/instruction/courses/LBC721/lecture_5/lecture_5.html
34.
The Difference Engine, Science Museum
http://www.sciencemuseum.org.uk/collections/exhiblets/babbage/diffeng.asp
35. The Difference Engine No. 2, Science Museum
http://www.sciencemuseum.org.uk/collections/exhiblets/babbage/diffeng2.asp
36. The Hare and The Tortoise, Science Museum
http://www.sciencemuseum.org.uk/collections/exhiblets/babbage/oldvsnew.asp
37.
The
History of Computation
http://www.worldofdave.ca/coms380/index.html
38.
The Third Generation
http://informatics.up.ac.za/staff/hanli/third%20generation.htm
39. TOP500 Supercomputer sites, TOP500.org
http://www.top500.org/list/2002/11/
40.
Transistorized!,
1999
http://www.pbs.org/transistor/
41. Unknown source (research in late 1980s or early 1990s)