Whence came the IBM PC?
Obviously, the IBM PC came from IBM, but that simplistic statement hardly tells the story of how IBM’s PC—introduced in 1981—changed the way people use computers. To celebrate the 20th anniversary of the IBM PC, T&MW dug into the past to put the PC’s developments into perspective. (See Acknowledgements.)
Few developments take place without acknowledging past work, and the IBM PC was no exception. For the IBM PC, the story began over 25 years ago—several years before its official introduction. In the mid ’70s, relatively inexpensive memory and microprocessor ICs made it possible for knowledgeable electronic experimenters to build their own computers. Prior to that, the smallest computers—minicomputers—took the space of a large microwave oven, cost thousands of dollars, and found use only in special applications.
|Figure 1. The IMSAI hobby computer looked more like a minicomputer from the 60s than what we think of as a PC. Yet this and other computers proved popular with hobbyists and experimenters and helped make IBM's PC a success. Courtesy of Thomas Fischer.|
But by 1975, commercial “hobby computer” kits such as the MITS Altair 8800 and the IMSAI 8080 (Figure 1), provided a basic front panel of lights and switches, a power supply, and a motherboard. The kit suppliers also offered a range of CPU, memory, and I/O boards. The resulting “systems” looked much like minicomputers. Hardy souls programmed these hobby computers—no one then called them personal computers—in assembly language. Some hobbyists, and a few people who saw business uses for small computers, wrote programs using a BASIC interpreter developed by a small software company called Microsoft, then based in Albuquerque, NM, the home of MITS.
In the late ’70s, hobby computers evolved from boxes with lights and switches to sleek packages that came with a keyboard and built-in software. Apple Computer offered its first commercial computer—the Apple II—in 1976, which Radio Shack answered the next year with its TRS-80. Both computers provided a built-in BASIC interpreter and let people store programs and data using an audio-cassette tape recorder. Later, both companies offered 51/4-in. floppy disk drives that simplified program and data storage. VisiCalc—a basic spreadsheet program—turned into the “killer application” that made businesses take these small “home” computers seriously.
IBM takes notice
The changing market for home computers didn’t go without notice at IBM. Unfortunately, the company had little experience with anything but mainframe computers, and its few forays into the small computer marketplace ended in routs. IBM introduced the 5100 in 1975 as an intelligent programmable terminal. The $9000 unit could operate as a stand-alone computer, and it could run BASIC and APL (A Programming Language) programs, but few people gave it any attention.
In 1980, IBM seemed more open to move into new businesses, so it encouraged new ideas within the company. William C. Lowe, the lab director of IBM’s Entry Level Systems (ELS) unit in Boca Raton, FL, had some ideas he wanted to try. The ELS unit worked on “low-end” computers, although not low enough to reach the level of a home computer. Lowe, who started his career at IBM as a test engineer, had studied the hobby- and home-computer markets. As he watched Apple and Radio Shack gain adherents who used their computers for business applications, the time seemed right to push IBM into the small-computer market.
In July 1980, Lowe presented his vision of a small computer to IBM’s corporate management committee. He saw an opportunity to compete in the market, and he believed IBM’s reputation for service and support would win it customers. At first, Lowe proposed IBM simply repackage a computer from a Japanese company, but the committee wasn’t happy with such an approach. So, Lowe changed course midway through his presentation and vowed to return in a few weeks with an alternate proposal (Ref. 1).
At the next meeting, Lowe proposed an IBM-originated design, and the committee gave its blessing to proceed slightly further. Lowe was impatient and promised to come back with a working prototype within a month—a seemingly impossible goal. His engineers at ELS had gained experience on a project called Datamaster, and that much-delayed project provided a place to “steal” talent and circuitry.
The goal of the Datamaster project was to build an all-in-one system containing dual 8-in. floppy disks, a keyboard, and a small display. The Datamaster wasn’t designed to work as a stand-alone computer, but many of the basic ideas could translate into Lowe’s prototype. In fact, the prototype Lowe delivered to the management committee simply repackaged a Datamaster’s circuits, which used an 8-bit Intel 8085 as its CPU.
The committee sees a prototype
|Table 1. Original suppliers of IBM PC components|
|8088 microprocessor and peripheral chips—Intel
5¼-in. floppy disks—Tandon (Note 1)
Power supply— Zenith (Note 2)
Printed circuit board—SCI
Keyboard—IBM (Lexington, KY)
Monitor— Various Asian manufacturers
1. Western Digital acquired Tandon in 1998.
2. Zenith sold its computer business in 1989 to Groupe Bull, which later sold it to Packard Bell NEC. Packard Bell no longer exists.
In August 1980, Lowe presented the prototype to the IBM management committee. His demonstration mainly displayed some graphic images, but the committee gave its consent to proceed with the project. And in a “first” for IBM, it approved the use of off-the-shelf components and assemblies from outside vendors for the new computer’s innards. IBM would eventually rely on outside suppliers (Table 1) for almost everything in the computer. Lowe now had approval for a formal project, dubbed Project Chess, and his goal was to develop a desktop computer code-named Acorn.
When it came time to design the Acorn, the designers—like engineers everywhere—stuck with what they knew. They had experience using Intel’s 8085, an upgrade of the 8080 used in many hobby computers, so the team gravitated toward Intel’s latest processor, the 16-bit 8086. Intel also supplied an 8088 chip that provided the same internal 8086-type CPU, but with an 8-bit external bus. The chips cost about the same—several hundred dollars each—but the 8-bit version would reduce the cost of I/O connections, support logic, and memory, so it became the chip of choice. In an article in Byte magazine, David J. Bradley, who wrote the Acorn’s basic input-output system (BIOS), cited four main reasons for the selection of the 8088 (Ref. 2):
1. The computer had to offer more than the maximum of 64 kbytes of memory available in most (then) current home computers. Thus, the design required a 16-bit processor. Although the 8088 supplied only an 8-bit external data bus, it also provided a 20-bit address bus (220) that could access 1 Mbyte of memory.
2. The processor and peripheral chips had to exist. IBM did not have time to develop custom devices.
3. The engineers had to be familiar with the CPU and its support chips.
4. An operating system and applications software had to exist for the CPU.
Software bolsters Intel
In addition to a CPU chip, Intel also could provide software that would translate 8080 or Z80 code so it would run on an 8086-family CPU. The hobbyist- and home-computer world was awash in 8080 and Z80 software that IBM might scoop up for its computer. So, IBM and potential software developers wouldn’t have to start writing code from scratch. (Intel also supplied native development tools for the 8086 CPU.)
According to Bill Sydnes, one of the engineers on the Project Chess team, another reason existed for choosing the 8088. The 8086 was deemed too powerful, and a computer built around it might compete too well with other IBM computers. But David Bradley says the 8086 offered only a twofold performance advantage, so there was no way an 8086-based computer would compete with other IBM products.
The designers of the IBM prototype computer quickly took advantage of the 8-bit support chips that Intel made for the 8088 CPU. The prototype used an 8237 direct-memory-access (DMA) controller, an 8259 8-level interrupt controller, and an 8253 counter/timer. Without these devices, the logic circuitry would have grown excessive, threatening the design goals for the system.
The Acorn prototype would also accommodate Intel’s 8087 math coprocessor, a chip that would perform floating-point math operation at high speed, thus reducing the processing burden placed on the 8088 CPU. Intel didn’t have the 40-pin 8087 chip ready when the IBM engineers produced the prototype, but they wired a socket for it anyway. As a result, the production IBM PC came with an empty 8087 socket for a user-installed math chip.
The computer Lowe planned to build would supply 16 kbytes of RAM on the system board (motherboard), along with 40 kbytes of ROM. The computer would use readily available 16-kbit dynamic RAM (DRAM) chips. A design using the 16-bit 8086 would double the needed memory chips—thus raising the cost of a basic computer. The ROM address space would provide 8 kbytes for the computer’s BIOS and 32 kbytes that would contain BASIC. The Radio Shack TRS-80 and Apple II already came with their own versions of BASIC built in.
David Bradley’s article in Byte also explained the 640-kbyte limit on the PC’s RAM. Of the 8088’s 1-Mbyte memory space, the designers reserved the upper 128 kbytes for ROM on the system board, 128 kbytes for video memory on display cards, and 128 kbytes for ROM and RAM on other types of adapter cards. Subtract those three 128 kbyte blocks from the 1-Mbyte space, and you have 640 kbytes left.
Lowe taps Estridge
Lowe had to move quickly from the small prototype team to a formal development team for the Acorn. And he needed a top-notch manager for the growing team. His first choice was Philip “Don” Estridge a likable leader who had been with IBM since 1959 and who was now located near Lowe in Boca Raton. Estridge took over leadership of the Project Chess team as Lowe moved into a vice president’s job elsewhere with IBM.
As it turns out, Estridge’s name, more than any other, became attached to the development of the IBM PC. When IBM presented someone to talk about the PC, it was almost always Estridge. In the early ’80s, if the IBM PC came up in conversation with engineers or hackers, so did “Don Estridge.”
The saying goes that success has many parents, but failure is an orphan. Thus, it proves difficult to get a definite list of people on the team that developed the IBM PC. “Team” could refer to the engineers who originally worked on the project in September 1980, members of a small task force that looked into the feasibility of a small computer, the inner core of Don Estridge’s group that worked in Boca Raton, FL, or perhaps some combination of all these groups.
Some sources say that IBM was innovative for making the Acorn an open-architecture computer. Actually, IBM did nothing new. The technology of home computers held few secrets. In fact, most of the growth in the home computer market came about because “open” computers were easy for users to modify with third-party hardware and software. And open computers made it easy for these third parties to engineer their products. As soon as a manufacturer offered a new computer, an entrepreneur was selling something for it.
The Apple II provided several expansion slots for disk-drive controllers, color graphics cards, and so on, and the TRS-80 furnished a bus connector and an external expansion chassis. Hobbyists were used to computers with expansion slots, and IBM simply followed the trend and provided five expansion slots on its Acorn motherboard.
Expansion slots weren’t new at IBM, either. The Acorn’s 62-pin edge connector came from the Datamaster product. And IBM’s engineers already had designs for several 8-bit add-in Datamaster cards they could adapt for the Acorn. According to David Bradley, the Acorn’s engineers changed only five of the Datamaster’s original signal definitions. In retrospect, though, Bradley said the designers failed to allot enough I/O addresses for all the cards that people eventually developed for the PC.
Software provides the key
At the same time hardware designers poured over schematics, some of their colleagues were searching for software to both run the Acorn and to run on the Acorn. By now, hobbyists who used computers based on Intel’s 8080 and on the Zilog Z-80 routinely used Microsoft’s BASIC and an operating system (OS) named CP/M. To IBM, BASIC and CP/M seemed worth a close look.
The CP/M operating system came out of work done by Gary Kildall, a consultant who worked for Intel in the early ’70s. While working at Intel, Kildall interfaced an 8-in. floppy disk drive to an Intel development system. To make the disk easy to use to save and retrieve information, Kildall wrote a program he called Control Program/Monitor (CP/M). (Some people say the M in CP/M stands for microcomputer or microprocessor, but the software’s manual calls the program a “monitor control program.”)
Kildall offered CP/M to Intel, but the company suggested Kildall market it on his own. So, he set up Intergalactic Digital Research, later shortened to Digital Research, to sell versions of CP/M to computer hobbyists. Eventually, Kildall rewrote CP/M so it relied on a separate, small section of code that he or the computer manufacturer could quickly customize for different computer models. As a result, the concept of a machine-dependent BIOS was born.
IBM wasn’t the only company interested in 16-bit CPUs. One such company, Seattle Computer Products (SCP) put an 8086 kit on the market in early 1979 and approached Kildall about rewriting his CP/M software for it. Kildall had other work under way and put off working on an OS for the 8086. Out of need, SCP hired Tim Paterson to write its own version of CP/M for the 8086. Paterson’s resulting OS went by the name QDOS, for quick-and-dirty operating system. In 1980, the company shipped a new version, named 86-DOS. Microsoft, now located in nearby Bellevue, WA, saw the SCP computer as a new outlet for its BASIC, so it rewrote the popular language to run on the 8086-based computer.
IBM was well aware Digital Research could offer help with an operating system, and Microsoft could supply a version of BASIC for the 8086. So IBM approached both companies. Gary Kildall, it turned out, knew a lot about programming, but not much about running a company. For whatever reason—and legends offer many—Kildall let an opportunity to work with IBM slip away. (See, “History buffs beware.”)
IBM visits Microsoft
When IBM approached Microsoft, the small company proved more receptive than Digital Research to the large company’s overtures. IBM wanted BASIC for its Acorn, and Microsoft could supply it. During talks between the two companies, conversation turned to the subject of an operating system for the Acorn. Bill Gates volunteered that in addition to BASIC, his small company could provide an OS for the 8086 or 8088 CPU. The IBMers liked what they heard. At the time, Microsoft didn’t have an OS for the 8086, but it knew a company that did—Seattle Computer Products.
|Figure 2. IBM's designers used wire-wrap breadboard techniques to produce several prototypes. The prototype shown here—in photographs of the top and bottom—is similar to the "computer" given to Microsoft for its development of MS-DOS and BASIC. Courtesy of IBM.|
Through several business arrangements, Microsoft licensed and then purchased outright the exclusive rights to 86-DOS. To let Microsoft test its software IBM supplied it with a wire-wrapped prototype of the Acorn (Figure 2) in November 1980. By early 1981, Microsoft was able to make 86-DOS run on the prototype, and in July of that year, the company adopted the name MS-DOS for its operating system and delivered final versions of the OS and BASIC to IBM. (IBM called its slightly different version of the operating system PC DOS.) But Microsoft’s future was not foretold. IBM had a habit of killing projects, even large projects, at the last minute, so Gates’ company had taken a big risk by tying its future so closely to IBM. Likewise, IBM had bet the success of the PC’s software on a 24-year-old entrepreneur.
Figure 3. The original IBM PC didn't look much different from other computers at the time, but the IBM brand, the availability of software, and IBM's complete documentation made the PC a hit among business, scientific, and engineering users. Courtesy of IBM.
|Figure 4. The original PC offered two connectors as its only built-in ports: one for the keyboard and one for a cassette recorder.|
The PC arrives
In New York City at 10:30 a.m., on August 12th, 1981, IBM announced, “. . . its smallest, lowest-priced computer—the IBM Personal Computer,” in a typewritten press release. The computer (Figure 3 ) would sell at retail in Sears and Computerland stores—another first for IBM—in its most basic form for $1565. The computer came complete with 16 kbytes of RAM, “cassette” BASIC in ROM, a 25-line 80-character/line display card, and little else ( Figure 4). IBM provided several configurations, including one with 64 kbytes of RAM, two 51/4-in. floppy disk drives, and an Asynchronous Communications Adapter (serial port) for $3045. Add a monochrome monitor for only $345. The computer ran at a blazing 4.77 MHz.
At the same time IBM announced the Personal Computer, the company let potential buyers know it had software, too. The release mentioned VisiCalc, office software from Peachtree Software, EasyWriter, Microsoft’s Adventure game software, and a no-frills communications program. IBM also noted it hoped to offer an 8086 OS from Digital Research, but CP/M-86 cost several times what Microsoft charged for MS-DOS, and even though some users thought it superior to MS-DOS, Kildall’s CP/M-86 never caught on.
Manual goes unmentioned
One of the best parts of the IBM PC—its Technical Reference manual—went unmentioned in the press release. This manual provided a wealth of information that exposed the PC’s hardware and software to scrutiny by developers. Release of such a cache of data was unthinkable for a company based on closed systems and proprietary hardware and software. Yet the Technical Reference manual provided 362 pages that laid bare the PC, from 82 pages of assembly-language BIOS listings to 50 pages of schematics. By using the Technical Reference as a guide, almost anyone with a grasp of software and digital electronics could produce an add-in board for the PC. And many people did just that.
Although little new existed in the PC, the computer made a lasting impression that continues today. You still can buy computers with one or two standard 62-pin expansion slots—now called Industry Standard Architecture (ISA) bus slots. The combination of having a computer backed by IBM, with open hardware and software, and the availability of third-party software that ran on the machine hit the market at just the right time.
Although IBM tried many times to duplicate its success with the original IBM PC, the company had subsumed the “rebel” designers and marketers into the corporate hierarchy. In the process, the company crushed the free and innovative atmosphere that led to the PC in the first place. Rules and procedures soon governed product development, and projects bogged down. In March 1983, IBM followed the PC with the PC XT, a computer that provided more memory in its base configuration and the capability to handle a 10-Mbyte “Winchester” hard drive. IBM followed the XT with the star-crossed PCjr, the 3270 PC, the IBM Portable Computer, the IBM PC AT, the IBM RT PC, and a list of forgettable systems that never caught the imagination of computer users the way the original PC did. Perhaps the PC was simply an aberration for IBM—a project that got out of hand and was lucky to succeed before management caught on. T&MW
1. McCartney, Laton, “Teaching the Elephant to Tap Dance,” White Paper 3.
2. Bradley, David J., “The Creation of the IBM PC,” Byte, September 1990. pp. 414–420.
This article relied heavily on the book, Blue Magic: The People, Power and Politics Behind the IBM Personal Computer, by James Chposky and Ted Leonsis (Facts On File Publications, New York, NY. 1988). For the sake of clarity, I have not noted every reference to materials taken from the book.
Thanks also go to Raymond Gorman and David J. Bradley at IBM for answering questions and providing information for this article.
Jon Titus has written real-time software and designed embedded systems and computer/instrument interfaces. He worked in electronics for 10 years and spent nine years at EDN magazine prior to joining T&MW in 1993. He has a BS from WPI, an MS from RPI, and a PhD from VPI.
History buffs beware
|Confusion exists about some aspects of IBM's PC development project, and the confusion will only grow as tales pass into folklore, and as memories blur. Most writers, for example, describe development of the PC as taking place in Boca Raton, FL, and they describe the use of the Intel 8088 processor from the start. Yet in his book, Inside Intel, Tim Jackson reports the PC project was run out of Austin, TX, and the PC was to be based on a Motorola chip (Ref. A). According to Jackson, Motorola ran late with its chip design, so the PC designers switched to the Intel 8088.
IBM did have a computer facility in Austin, but according to David Bradley, the Austin group worked on IBM's DisplayWriter, an 8086-based system. The Austin operation approached desktop computers from the perspective of large-business users, and it was not involved with the development of the PC. Bradley also says IBM had many small projects under way, but Project Chess was the only “small computer project with more than 10 people working on it.”
Several versions of the Digital Research/Gary Kildall story exist. In one version, Kildall failed to show up for the meeting with IBM officials because he was flying a private airplane either for amusement or on a return trip from visiting a client. A second version of the story has Kildall flying in slightly late to meet the IBM representatives. After extensive talks, IBM and Kildall spar over what Kildall's company, Digital Research, will charge IBM. The deal falls through. In yet a third version of the story, Kildall left working with hardware companies to his wife Dorothy, who had little time to meet with IBM people the day before she was going on vacation. Confirming such stories proves difficult. Gary Kildall died in 1994, almost unnoticed by the computer world, and Kildall's ex-wife, Dorothy McEwen, refused my request for an interview. Thus, several versions of the story live on. Novell acquired the assets of Digital Research, and Caldera acquired those rights in 1996.
Gordon Eubanks, president and CEO of Oblix (Cupertino, CA), and a former colleague of Kildall, said, “The real issue wasn't that Gary refused to talk to IBM. The real issue was that Microsoft had a much better vision for the business” (Ref B).— Jon Titus
A. Jackson, Tim, Inside Intel: Andy Grove and the Rise of the World's Most Powerful Chip Company, Penguin Putnam, New York, NY, 1997.
B. MaxFrame Corporate Recollections of Gary Kildall.