HP: The Accidentally, On-Purpose Computer Company
Hewlett-Packard entered the pantheon of computer manufacturers through a side entrance. It’s pretty clear from historical documents and interviews with people who were present at the time that Bill Hewlett and Dave Packard held diametrically opposed views on the computer business through most of the 1960s. At the beginning of the decade, HP was clearly nothing but an instrument company—a very successful instrument company with a wide and growing product line. Yet Packard observed that something truly important was occurring in the world of data processing (the 1960s term for IT or information technology) in the form of the minicomputer and he didn’t want HP to needlessly miss out on a profitable market opportunity. He even made a shopping trip to the Digital Equipment Corporation (DEC), located in Maynard, Massachusetts, early in DEC’s history but founder Ken Olsen either wanted too much money for his company ($25 million) or otherwise could not come to terms with Packard.
(Note: HP “merges” with Compaq about 40 years later and therefore finally succeeds in buying DEC. Compaq had previously ingested the remains of DEC’s rotting carcass, which it salvaged from the immense pile of dead and dying minicomputer companies that also included Prime, Data General, Computer Automation, General Automation, and Sperry-Varian. All of these formerly successful companies withered away as a result of IBM’s introduction of the PC in 1981, which demolished the minicomputer market.)
Clear-eyed Bill Hewlett saw that the business data-processing market was already overcrowded back in the early 1960s. Nevertheless, HP did enter the minicomputer market in 1966, along with Varian Data Machines. Over the next two years, General Automation, Computer Automation, Data General, Micro Systems, and Lockheed entered the fray.
The story of how HP finally did enter the computer market involves a bit of Dave Packard subterfuge, a phantom HP subsidiary sporting an upside-down HP logo, the acquisition of a minicomputer company owned by an unlikely parent named Union Carbide, and a team of freshly hired, sandal-shod, long-haired, hippie-freak programmers who refuse to wear ties and work normal business hours.
You say DYNAC and I say Dymec
In 1956, Hewlett-Packard formed a wholly-owned subsidiary named DYNAC, Inc. with financing provided by contributions from HP executives. Depending on whose story you believe DYNAC was:
- A financially advantageous side deal among top HP executives
- A way to take on projects HP would not ordinarily take, such as missile test systems, without negatively affecting the employee profit-sharing program
In any case, the name DYNAC was selected because the first two letters, “dy,” looked a lot like the lower-case letters “hp” turned upside down. Consequently, DYNAC’s logo was simply HP’s logo inverted. This move obviously saved some dollars on developing a new corporate identity. However, Westinghouse owned a DYNAC trademark so HP’s DYNAC subsidiary became DYMEC, Inc in 1958 (preserving the essential letters “d” and “y” in its name). DYMEC, Inc. ceased to be a subsidiary and became the Dymec Division of HP in 1959 and then becomes the Palo Alto Division in November, 1967.
At first, DYNAC/DYMEC/Dymec Division made components for the HP manufacturing divisions. However, at some point, DY*** became a sort of HP systems integrations house. It created test systems using HP test equipment and it designed and built any extra gear needed to complete these systems. As part of that charter, Dymec investigated the use of minicomputers as instrument controllers. Specifically, it experiments with DEC’s PDP-5 and PDP-8 minicomputers in 1964 and considers the specialized software and custom-designed interface hardware that would be needed to turn DEC’s minicomputers into instrument controllers. Dymec concludes that it needs a machine with more flexibility than currently available minicomputers.
Dymec + DSI = Cupertino
Meanwhile, having possibly found the fledgling DEC too expensive a purchase or realizing that Ken Olsen will always be too difficult to deal with, Dave Packard shops around for another minicomputer company to buy. He finds a very small and reasonably priced 5-person firm named DSI (Data Systems, Inc.) in Detroit. It’s owned by Union Carbide. When asked why Union Carbide owned a minicomputer design firm, HP Labs manager Barney Oliver replied, “We didn’t demand an answer to that question.” Apparently some executive at Union Carbide had a diversification brainstorm that didn’t pan out.
However, Bill Hewlett refuses to agree to the DSI purchase until Packard reframes DSI’s design as an “instrument controller” instead of a “minicomputer.” Ken Olsen pulled the same subterfuge on the US Government shortly after he founded DEC in 1957. Olsen discovered that Congress had passed a law against the purchase of any more computers until the ones the governmant already owned were 100% utilized. Nevertheless, a large division of the government needed to collect and analyze “seismic” data from “earthquakes” and therefore readily agreed to the purchase of DEC’s “programmed data processors,” which is precisely when and why the abbreviation PDP appeared on DEC’s minicomputers for the company’s first two decades.
HP buys DSI in 1964 and with it gets the rights to a computer design called the DSI 1000. Four of the five DSI employees end up joining HP, which forms a new computer division by blending former DSI employees, some of its Dymec Division employees, and other employees from its instrumentation divisions. The new division later becomes the Cupertino Division when it’s installed in a building HP purchases from Varian in Cupertino.
Kay Magleby’s interest in the new technology of digital integrated circuits will lead him to become the architect of HP’s first minicomputer. (Photo courtesy of the Hewlett-Packard Company.)
Initial work on HP’s first minicomputer starts in a corner of the Dymec Division building under Kay Magleby whose previous HP design experience was with sampling oscilloscopes and a vector voltmeter. Magleby had become interested in working with digital product designs and the new digital integrated circuits that had started to appear on the market. He and Paul Stoft had developed the concept of an HP instrumentation controller at Dymec during 1964. The combination of DSI’s minicomputer-design expertise with Magleby’s and Stoft’s carefully considered goals for an instrumentation-oriented minicomputer and Dymec’s systems know-how allows the design and release to manufacturing of HP’s first minicomputer, the HP 2116A, to proceed at a rapid pace. (Stoft will become manager of HP’s audio-video division.)
The resulting processor instruction-set architecture (ISA) is a model of efficiency and the HP 2116A processor shares some attributes of RISC architectures that will surface nearly 20 years later:
- There are only 68 instructions
- All instructions are 16 bits wide (fixed instruction width)
- All instructions execute in one memory cycle (1.6 microseconds)
However, in many ways, the HP 2116A ISA is quite un-RISC-like:
- It’s an accumulator-based machine with instructions that operate on data in memory
(the opposite of a RISC machine’s load/store architecture)
- It’s a 2-accumulator machine
(RISC processors usually have many general-purpose registers that serve as accumulators)
- Its register-reference instructions can execute multiple operations per instruction
(RISC processors generally execute only one operation per instruction)
In about another five years, the HP 2116A’s instrumentation-oriented, I/O-intensive minicomputer architecture will serve HP’s desktop calculator and computer projects extremely well. Although it will not be used for the first-generation desktop calculator, the HP 2116A’s ISA will serve as the foundation for HP’s extremely successful second- and third-generation desktop calculators and computers.
Counterculture software engineers
Roy Clay leads the software development team for the HP 2116A. One of his philosophies is that the software must be ready with the hardware on the date of introduction and he’s willing to do whatever is necessary to make that goal a reality. Many members of his programming team are new graduates from Stanford’s computer science program. While the “real” engineers at the Dymec Division must wear regulation white shirts and ties (like all professionals of the day) and they are required to wear steel-toed shoes if they walk onto the manufacturing floor, Clay’s programmers prefer sandals, wear sport shirts, work at odd hours, and generally behave the way programmers have worked since computers were invented. They’re moved to a trailer with a leaky roof in the parking lot so as not to disrupt the other HP employees and they eventually end up in a basement with a separate entrance because of their odd working hours.
Hewlett-Packard introduces the HP 2116A minicomputer at the San Francisco Fall Joint Computer Conference in November, 1966 and discloses technical details of the machine in the March, 1967 issue of the Hewlett-Packard Journal. HP claims that the HP 2116A is the industry’s first 16-bit minicomputer but it wasn’t. Although the HP 2116A predates DEC’s hugely successful, 16-bit PDP-11 by some four years (the PDP-8 was a 12-bit machine), the DDP-116 minicomputer designed by Gardner Hendrie (who would subsequently join Data General and design the microNova), and introduced by the Computer Control Company (later bought by Honeywell) in April of 1965, was actually the first commercial 16-bit minicomputer to hit the market.
HP’s 2116A is the company’s first product to incorporate digital integrated circuits in its design. Significantly, the HP 2116A’s software (FORTRAN compiler, assembler, linker, loader, operating system, and I/O drivers) and hardware are both ready at the time of introduction, which is not the norm for the computer industry of the day. In general, computer vendors would roll out the hardware and the software would follow some time later.
It’s not a computer, it’s an instrument controller
Paul Ely, Jr., then the engineering manager at HP’s Microwave Division wrote a short preface to Magleby’s technical article about the HP 2116A in the March, 1967 issue of the Hewlett-Packard Journal. In that preface, Ely calls the HP 2116A “an unusual new instrumentation computer” and that’s precisely what the HP 2116A was. In addition to all of the usual attributes of a mid-1960s minicomputer, the HP 2116A had several features that made it especially useful as an instrumentation controller.
First and foremost, the HP 2116A had an oversized cabinet with 16 empty card slots that would accept a wide variety of interface cards. Add-on modules could raise the machine’s maximum number of interface cards to 48.
At its introduction, HP engineers had designed interface cards that connected the HP 2116A to more than 20 HP instruments including “counters, nuclear scalers, electronic thermometers, digital voltmeters, ac/ohms converters, data amplifiers, and input scanners.” These were the earliest HP instruments with digital interfaces and the HP 2116A’s introduction marks the start of the age of automated test systems.
Another series of interface cards connected the HP 2116A to “most kinds of input/output devices, such as magnetic tape recorders, teletypewriters, paper tape readers and punches, and dataphones [an archaic word for modems].” Hewlett-Packard engineers were already busy designing additional HP 2116A interface cards for printers, card readers, and card punches. These additional cards would be introduced later as would another essential minicomputer peripheral, a disk drive.
The HP 2116A was clearly designed as an interface engine, much as it was first envisioned by Magleby and Stoft. In addition to the large number of card slots devoted to I/O, its internal architecture further aided I/O through hard-wired, prioritized interrupt vectors. Each of the HP 2116A’s interface slots had a permanently assigned, hard-wired interrupt-vector location in the minicomputer’s memory space. An interface card that asserted its interrupt signal could activate the appropriate interrupt service routine in microseconds (assuming its priority was the highest active interrupt at the time).
Other minicomputers of the day used polled I/O software routines and might require a millisecond or more to initiate an I/O transfer after the hardware made a service request. Millisecond response times might be fine for buffered peripherals like printers, but they’re far too slow for real-time data-acquisition systems. If even faster I/O was required, the HP 2116A would soon support direct-memory-access (DMA) I/O transfers at 1.2 Mbytes/second, which was 10,000 times faster than it’s conventional transfer rates.
However, in 1966, there’s still a relatively small market for instrumentation controllers and there is, in fact, a much bigger market for business data processing. This is precisely the market that business-savvy Dave Packard had smelled back in 1964. In short order, HP discovers that it’s selling far more HP 2116A minicomputers for business applications than it is for instrumentation applications despite the company’s best efforts to the contrary. Bill Hewlett restrains his instrumentation bias in the face of the minicomputer’s crushing success in the data processing market and he will eventually start to use an HP 2100 series minicomputer to keep track of bull breeding stock that he and Dave Packard run on their Little Basin ranch in California.
As a result of its success in the business-computing market, HP re-engineered the HP 2116A and introduced cost-reduced versions of the HP 2116A called the HP 2115A and HP 2114A minicomputers in 1967 and 1968 respectively. Without the instrumentation-slot bays, these new machines are much smaller and are therefore less expensive. The HP 2114A and HP 2115A launch the Cupertino Division into a long series of successful HP 21xx and HP 21MX minicomputers that HP will not be able to kill despite five serious attempts to introduce successors.
By 1978, HP is the fourth largest minicomputer manufacturer, trailing only DEC, IBM, and Data General. The 16-bit HP 21xx architecture will live for more than 20 years, making HP one of the minicomputer vendors that truly succeeded, in spite of Bill Hewlett’s initial reluctance at entering the market and the company’s subsequent, narrow focus on instrumentation control (until the market forced it to broaden its horizons).
However, the summer of 1965 has ignited another round of developmental brainstorming in Barney Oliver’s R&D lab. Two unrelated but fortuitous visits by wandering computational visionaries seeking homes for their radical ideas have planted the seeds of a desktop-computer revolution. The resulting product will be wonderful but quite strange—so strange in fact that the “maverick” Cupertino Division and its counterculture engineers will want nothing to do with the first-generation machine spawned by this work. Thus begins the story of how HP’s instrument division in Loveland, Colorado got the HP 9100 desktop calculator.
Source materials for this Web page about the HP 2116A minicomputer include:
Todd Poyner, “25 Years of Real-Time Computing,” REAL-TIME Interface, An Interex Publication, August, 1991, available at www.interex.org/tech/csl/RTE/archive/poyner1.htm.
Kip Crosby, “HP’s Early Computers, Part One: An Interview with Barney Oliver,” The ANALYTICAL ENGINE, Journal of the Computer History Association of California, Volume 2, Number 3, May 1995.
Leslie Jaye Goff, “A new machine, a new way,” Computerworld, May 3, 1999.
Leslie Goff, “1966: HP’s radical move,” Computerworld, July 6, 1999.
Paul Ely, Jr., “The Engineer, Automated Network Analysis and the Computer—Signs of Things to Come,” Hewlett-Packard Journal, March, 1967.
Kay B. Magleby, “A Computer for Instrumentation Systems,” Hewlett-Packard Journal, March, 1967.
Donald P. Kenney, “MINICOMPUTERS, Low-Cost Computer Power for Management,” AMACOM, 1978.
David Packard, “The HP Way, How Bill Hewlett and I Built Our Company,” HarperCollins, 1995.
Kenneth Jessen, “How it all began, Hewlett-Packard’s Loveland Facility,” J. V. Publications, Loveland, Colorado, 1999.