M•I•C•R•O P•R•O•C•E•S•S•O•R E•V•O•L•U•T•I.O•N Reprinted by permission from BYTE, September 1985.. a McGraw-Hill Inc. publication. Prices quoted are in US S. EVOLUTION OF THE MICROPROCESSOR An informal history BY MARK GARETZ Author's note: The evolution of were many other applica- the microprocessor has followed tions for the new memory a complex and twisted path. To chip, which was signifi- those of you who were actually cantly larger than any that involved in some of the follow- had been produced ing history, 1 apologize if my before. version is not exactly like yours. About this time, the The opinions expressed in this summer of 1969, Intel was article are my own and may or approached by the may not represent reality as Japanese calculator manu- someone else perceives it. facturer Busicom to pro- duce a set of custom chips THE TRANSISTOR, devel- designed by Busicom oped at Bell Laboratories engineers for the Jap- in 1947, was designed to anese company's new line replace the vacuum tube, of calculators. The to switch electronic sig- calculators would have nals on and off. (Al- several chips, each of though, at the time, which would contain 3000 vacuum tubes were used to 5000 transistors. mainly as amplifiers, they Intel designer Marcian were also used as (led) Hoff was assigned to switches.) The advent of assist the team of Busi- the transistor made possi- com engineers that had ble a digital computer that taken up residence at didn't require an entire Intel. Hoff looked at the room full of vacuum Busicom design and de- tubes, relays, and special cided it was too complex air conditioning. Now a to be cost-effective. He computer would take up had worked before with only half a room and Digital Equipment Cor- operate much more poration's PDP-8 mini- quickly. computer, which had a It was not until 1959 that very small instruction set. engineers at Texas Instruments ing relatively few transistors), through He reasoned that much of the figured out how to put more than one MSI (medium-scale integration, calculator's complexity could be transistor on the same material (called around 50 or more transistors), LSI reduced if they used a small general- the substrate) and connect them (large-scale integration, with thou- (continual) together without wires. Thus was born sands of transistors), to VLSI (very- Mark Garetz (Viasyn Corporation, 26538 the integrated circuit, or IC. 'Ibday large-scale integration, which can con- Danti Court, Hayward, CA 94545) is direc- these thin flat pieces of silicon can tain millions of transistors). tor of advanced projects at Viasyn. He has contain millions of transistors, and we In 1969, a year-old company named been a computer hobbyist since 1974, he de- call them chips. Intel announced a 1K-bit RAM chip. signed COmpuPro's 8085/8088 computer, Integrated circuits may range from There were not yet any microcom- and he is chairperson of the IEEE 696 SSI (small-scale integration, contain- puter chips to hook it to, but there Committee. ILLUSTRATED BY WILLIAM LOW DECEMBER 1985 • JUST COMP JTERS 261 MICRO PROCESSOR EVOLUTION purpose processor. Such a design, design and the project at Intel gave icant upgrade to the 8008 that re- using software rather than electronics birth to the 8008. quired only six support chips, had 75 to do the calculating, would greatly in- The 8008. introduced in April 1972, instructions and a tenfold increase in crease the memory requirements of was the first 8-bit microprocessor on throughput over the 8008, and ad- the calculator—but then, Intel was in the market. It required at least 20 sup- dressed 64K bytes of memory. (No the memory business. Hoff also real- port chips, but it had 45 instructions program, most people thought, could ized that this processor could be put that it executed at 300,000 instruc- ever be that large!) to other applications and he sold the tions per second, and it adddressed The 8080 design was proposed by idea to Intel management. a whopping I 6K bytes of memory. Faggin, but the design team was The Busicom engineers were still That was a lot of memory then, and headed by Masatoshi Shima, a young pursuing their original design when the 8008 was a considerable upgrade engineer Intel had wooed away from Hoff and his group started work on of the 4004. Busicom. Having learned from the their alternative design. And although The documentation for the 4004 limitations of the 4004 and 8008, the the Busicom engineers had simplified and 8008 was cryptic (at least it designers made improvements to their design, each chip still had over seemed so to me; I didn't have a com- make their new chip a truly useful 2000 transistors, and it would take 12 puter background at the time). The computing engine. The 8080, the first chips to make a working calculator. documentation assumed that you microprocessor not aimed at logic re- Hoff's team figured it would take 1900 knew what everything was before you placement, looked much more like a transistors to build their processor. started reading it (a still-common fail- computer than anything that had Hoff's general-purpose processor ing of technical literature). come before it, and it was much easier design was chosen over the Busicom Intel's primary goal with the 4004 to use from a hardware standpoint. design, and Intel got a contract from and 8008 was to replace "random The January 1975 Popular Electronics Busicom to produce the chip that logic'=-another way of saying "lots of magazine featured the first in a series later became known as the 4004. SSI and MSI wired together:' Few peo- of construction articles on the Altair Actually making the chip proved to ple thought that these chips were 8800, a so-called "minicomputer" be difficult until Federico Faggin (who suitable for general-purpose comput- based on the 8080. The Altair was de- later founded. Zilog) joined Intel in ing. But a few visionaries were in- signed by MITS (Micro Instrumen- early 1970. He took the chip from trigued by the possibility of owning tation and Telemetry Systems), which concept to silicon in just nine months. a computer that could actually do was founded by Ed Roberts as a vehi- At first Intel sold the 4004 exclusive- something. Mlle, computer kits had cle for supporting his experiments in ly to BusicOrn, but in the summer of been offered previously, but they electronics. The whole Altair kit, in- 1971, it gained the right to sell the were more useful for demonstrating cluding the 8080 processor, mother- chip set to other manufacturers. computer principles than for doing board, power supply, front panel with In November 1971 Intel advertised computing tasks. The availability of lots of lights, and 256 bytes (not 256K the 4004 as a four-bit processor that the 8008 changed all that. bytes) of memory sold for $395. performed 60,000 operations per sec- In 1973 Scelbi Computer Consulting People thought it was a misprint. ond. By February 1972 Intel had sold Inc. announced the first general- The 8080 chip, introduced just nine $85,000 worth of chip sets. purpose microcomputer based on the months before, had been selling for 8008. This was followed by the $360 all by itself. But MITS had made THE BIRTH OF 8-BITS RGS-008 from RGS Electronics. Then, a special deal with Intel, and the price At the same time the 4004 was be- in July 1974, Radio-Electronics magazine of the Altair was real. MITS sold more ing developed, CTC (Computer Tech- introduced Jonathan Titus's Mark-8 in computers in the first day than it had nology Corporation, now Datapoint) a series of construction articles. hoped to sell during the whole life of asked both Intel and Texas Instru- Until then, all computer articles and the product. ments to design LSI chips for a new ads had been confined to amateur- The Altair played a significant role intelligent terminal. Both companies radio publications. The Mark-8 was in the success of the 8080, largely proposed an 8-bit general-purpose the first computer to hit a general- because programmers now had a processor. Note the pattern develop- interest electronics magazine. These reason (and a good excuse) to write ing: 4-bits for calculators because they early microcomputers were still more software for a microcomputer chip. work in BCD (binary-coded decimal) demonstration tools than useful, but Also, the Altair's open bus architec- and 8-bits for terminals because they the small-computer revolution had ture (an improved version of which deal with ASCII characters. begun. later became the S-100/IEEE 696 bus) Interestingly. CTC chose neither allowed people to begin making pe- solution; it built its terminal with stan- THE MIGHTY 8080 ripherals for the computer. dard logic ICs. But TI and Intel went In April 1974 Intel changed the way One such peripheral was a disk con- ahead with their projects anyway. TI we think about computers forever. troller from Digital Microsystems that eventually got a patent on its chip They announced the 8080, a signif- featured the use of a new operating 262 JUST COMPUTERS • DECEMBER 1985 MICRO PROCESSOR EVOLUTION system for the 8080 called 6800 were both selling for CP/M (Control Program for $179 in single-piece quanti- Microcomputers.) CP/M, ty. I remember standing in brainchild of Naval Post- the lobby (actually a living graduate School instructor room) of E-Mu Systems Gary Kildall, sold for $70 with Scott Wedge and Dave and played a major role in Rossum, who had just de- the success of the 8080 and signed some 8080s into its architecture.