A microprocessor is a central processing unit (CPU). It can be defined as

The first microprocessor, the Intel 4004, was formally announced to the world in November 1971. At that time it had no market and there was little wish in Intel's marketing division to push the product. Only with the designers of the chip arguing that it would sell more memory was it finally accepted. In 1996, 25 years on, the principal products from Intel are microprocessors and Intel has a net income of $5,200,000,000.

Behind the production of the 4004 there are two key companies, Intel and Busicom. Intel Corporation was founded in July 1968 by Dr Gordon E. Moore and Dr Robert N. Noyce. The two had previously founded Fairchild Semiconductor, by then a major force in the industry. The new company was set up to develop some of the ideas for new types of memory products initially started by Fairchild and they were looking at the new process of producing a silicon gate MOS (metal-oxide semiconductor).

The second company involved was a Japanese firm, a young company looking to break into the calculator market, the first mass market for integrated circuits. Busicom searched for a company in America that could develop chips for a series of calculators ranging from small desktop models to teller machines. They company they chose to do this was Intel.

Three men from Busicom went to Intel with the initial designs for the calculator and the proposal was made by Mr Masatoshi Shima. Shima was an organic chemistry graduate who had used a mainframe computer at University for analyzing organic chemical compounds. He joined Busicom at a time when the new IC (integrated circuit) technology was just taking off and his lack of previous experience with larger transistors proved an asset.

The specifications for the calculator were based on a design Busicom had already successfully implemented using transistor logic. The design was for a decimal computer with a macro-instruction set in read-only memory and hardware of seven different LSI (large scale integration) chips and the logic schematics for the design were about 85% finished when Shima made his pitch.

Intel, however, were not in a situation to fully utilise the work done by Busicom. Intel was still a very small company and they had no logic or circuit engineers. Shima found it frustrating attempting to explain the logic schematics to Intel and this communication problem lead to an impasse lasting a few months, during which the project did not advance.

The Intel team leader for the project was Marcian 'Ted' Hoff, a former computer research associate at Stanford. He was assisted by a programmer, Stan Mazor. Hoff had been working with main frames that took only a small and simple built-in instruction set and could yet produce complex results with sophisticated programming. In reaction to the Busicom specifications, Hoff suggested the idea based on the main frames he had used but LSI architecture and with the calculator functions as programs. The Japanese, however, merely expressed agreement with the view that the initial specification was too complicated. They agreed to simplify it but did not want to produce something that wasn't geared towards calculators.

The conflict between the Busicom idea and Intel's had parallels in the industry in general. On the one hand, building specific circuits for each calculator allowed individual calculators to utilise most effectively the MOS technology. On the other, imitating large scale computers with chip architecture would allow general purpose chips that could be programmed for whichever calculator needed them. Intel's plan had always been to create proprietary products, so CPU on a chip was an idea that appealed to its founders. When Hoff took his idea to them, Moore and Noyce told him to develop the idea.

Hoff finally created a design that appealed to both Busicom and Intel. For Shima he drew a diagram with three blocks - a four-bit arithmetic unit, a four-bit general purpose register block and a twelve bit address stack. Included in the diagram were a number of I/O ports, for the keyboard and other peripherals, though Hoff told Shima he had no idea how to implement them. In a final management meeting, the two designs were pitched to Busicom and they agreed to go with new idea from Intel.

Even when Busicom accepted the Intel design there were a number of main problems which Shima and the Intel team had to work out. Hoff's initial basic design was workable but he had not considered real applications, particularily in designing the instruction set list. Firstly, there was a lack of comparable and useful LSI systems. This was resolved by the decision to design a whole family of chips - the 4000 family. Secondly, binary to decimal and vice-versa programs took a lot of memory, so a ROM conversion table was used. Thirdly, keyboard control had to be introduced. The fourth problem was how to have real-time control. This problem did not present itself with an immediate solution. The fifth problem was a lack of a register indirect jump. Adding an indirect addressing mode solved the fifth and, also, the fourth by allowing an external input to be one of the conditions in conditional branch instructions. Throughout, the combination of Shima's experience in decimal calculator design, Hoff's work with large mainframe computers and Mazor programming skills proved invaluable in creating a workable design.

Again Intel came across a problem due to lack of engineers. This one could only be solved by finding someone to design the actual chip. During the search for a suitable designer, no further work took place on the chip. The answer was found, after a number of months, in Federico Faggin, a employee of Fairchild Semiconductors, where he developed the original MOS silicon-gate technology. His arrival was very shortly followed by the return of Shima from Japan, who was very disappointed to find no further development on the 4004. Faggin immediately began working very long hours - up to 16 hours a day - to produce the chip design. For each chip the definition of the chip architecture and basic specifications were needed. These were completed by Hoff and Mazor, with Shima's help. Then came logic design, circuit design, layout, ruby cutting, mask making, wafer production and finally chip verification and debugging. All of this was labour intensive for Faggin who, discouraged by Intel from using their mainframe, worked with a slide rule, using graphical analysis.

At the end of 1970 the prototype 4000 family chips arrived at Intel. The first three - 4001: 2,048 bit ROM, 4002: 320 bit RAM and 4003: 10 bit input-output shift register all worked straight from the initial pressing. The first 4004 had one entire layer missing from fabrication but, after this mistake was rectified, Faggin could find no errors. The chips were sent out to Japan in March 1971. On receiving the 4004, Shima, despite having created numerous small test procedures, tested the whole calculator in front of a large audience of colleagues. It too worked first time.

By the time 4004 came out, however, Busicom were having financial difficulties. Faggin realised the potential in the 4004 as more than the main chip of a calculator and pushed for Intel to obtain the rights from Busicom. To persuade Intel's management of a wide field of uses for the processor, he used the 4004 as the controller in testers for the 4000 family range. The calculator price wars in Japan helped to force Busicom's hand and they renegotiated the price on the Intel contract, including handing over rights to Intel.

The decision to market the 4004 was still a hard one for Intel to make, as it was a product that there was no demand for other than that Intel had to drum up itself. Intel was given a further jolt when a Texas Instrument advertisement appeared containing a chip with the caption "CPU on a Chip", developed for Computer Terminal Corp, a company Intel had also been developing a chip for.

The TI chip turned out to be dead in the water, a clear relief for Intel. The 4000 series was launched and the chip they had developed for CTC they renamed the 8008. People began to expect a lot from these products and the minor problems encountered were dealt with. Shima joined Intel after a spell at Ricoh to aid the development of the next processor, the 8080. With this chip came the real arrival of functional and appreciated microprocessors in 1974. Within a year of its launch, it was designed into over one hundred different lines.

Microprocessors are now used by objects as far apart as phone cards and aeroplanes and there is no evidence to suggest that the increase in use will slow down. As more are used, so too do microprocessor abilities and speeds increase. The 4004 had 2,300 transistors on the chip. The 1995 Pentium has 5.5 million. The pace of this increase seems to follow a trend suggested by Dr Moore in 1965, before Intel was founded and the 4004 built. This is that capacity doubled every two years, a formula known as Moore's law. This trend has been followed since the first days of microprocessor's and there is no evidence to suggest that it slowing down after 25 years.

________________________________________________________________________________

Sources

IEEE Micro, Dec 1996, "Celebrating the Microprocessor"

Daniel R. McGlynn, Microprocessors Technology, Architecture, and Applications John Wiley & Sons, Canada, 1976 (a)

Michael S. Malone, The Microprocessor A Biography TELOS Springer-Verlag, New York, 1995

The Red Herring Magazine, Sept 1995, "The Accidental Entrepreneur" http://www.herring.com/mag/issue23/accidental.html

Interview with Masatoshi Shima http://www.ieee.org/history_center/oral_histories/transcripts/shima.html

Intel http://www.intel.com/

pc WEBOPAEDIA http://www.pcwebopaedia.com/