Intel Introduces the Pentium Processor

Date March 22, 1993

When Intel Corporation, the leading microprocessor company, introduced the Pentium chip, it set records in transistor microminiaturization and introduced superscalar architecture.

Locale Santa Clara, California

Key Figures

• Gordon E. Moore (b. 1929), cofounder and chairman of the board of Intel Corporation
• Andrew Grove (b. 1936), chief executive officer of Intel Corporation

Summary of Event

Intel Corporation was the original microprocessor company. Robert Norton Noyce, who founded the company with Gordon E. Moore, was one of the inventors of the integrated circuit. Intel employee Marcian Edward Hoff, Jr., invented the microprocessor when he recognized that the logical development of the integrated circuit was to place the entire circuitry of a computer’s central processing unit (CPU) onto a single piece of silicon. This discovery made the microcomputer revolution possible.

However, by the 1990’s, trouble was nipping at Intel’s heels. Although its 80486 microprocessor was one of its most profitable products ever, and even older processors such as the 286 and 386 continued to be strong profit centers, other chip manufacturers were encroaching on Intel’s lead. Apple Computer, International Business Machines (IBM), and Motorola had just announced the formation of a joint venture (known as the AIM alliance) to produce a new kind of microprocessor that would use reduced instruction set computing (RISC), a technology formerly seen only in high-end mainframes. Additionally, companies such as Cyrix and Advanced Micro Devices (AMD) were copying Intel’s leading chip designs and selling them under similar names, thus making significant inroads on Intel’s market share.

As a result, Intel’s chief executive officer, Andrew Grove, and several other key figures of the company decided on a radical step. Taking a chapter from Andrew Carnegie, the Gilded Age industrialist, who once tore down a new steel mill because innovation in steelmaking technology had made it obsolete, Intel would not leave the 486 chip to run its normal product life cycle. Instead, they would preemptively discontinue the product and replace it with a new processor of a substantially more advanced design.

Originally this chip was to be the 80586, following the numerical naming pattern of previous microprocessors in its line. However, the U.S. government refused to permit Intel to trademark a numerical designation, which left Intel vulnerable to the production of knockoffs by rival chip makers using such designations as i586. As a result, Intel took the radical step of giving the chip a name. A team did extensive market research to determine what syllables would best evoke the proper sense of an advanced microprocessor and finally settled on the name Pentium. “Penta” is the prefix for “five” and “ium” has a general science and technology association.

The Pentium was also technologically innovative. Not only would it contain a whopping three million transistors, but it was also the first to employ superscalar architecture. Unlike its RISC rivals, which followed the pattern of all previous microprocessors in having a single data pipeline and executing one instruction per clock cycle, the Pentium would have a dual data pipeline, which meant it executed two instructions each clock cycle in two separate execution units. The result was a chip that ran twice as fast with less complexity. Furthermore, it remained compatible with the previous chips of the X86 family because it continued to use the same Intel instruction set.

In October, 1994, a call came to Intel’s call center in Folsom, California, that revealed that all was not well with the Pentium. Dr. Thomas Nicely was a mathematician at Lynchburg College, Virginia, and he was using several Pentium-based computers for number crunching related to the search for very large prime numbers. Nicely’s problem was simple: When he asked the computer to perform a long division problem with two very large numbers, it returned a wrong answer. He checked all the obvious problems: the data, the program, and conflicting programs on the hard drive. He even went to a computer store and entered the data on a brand-new computer, but he received the same wrong result. There remained only one explanation: that the Pentium processor was faulty.

When Nicely attempted to report the problem to Intel, he received no help. Instead of recognizing the problem as a major one and referring it to the Pentium engineering department, the help desk employee trivialized the problem. The lack of attention given to the problem Nicely had discovered, however, was not entirely a matter of indifference or incompetence on the part of the help desk employee. The people in the call center had not been provided with an up-to-date list of known bugs in the Pentium processor, and this particular bug had been known for almost six months.

The bug involved a lookup table used by the chip’s floating point unit (FPU). To speed processing, the table had been encoded on the silicon of the chip itself, but 5 of the 2,048 entries were incorrect. Because of the way an FPU handled long division of very large numbers, errors would usually correct themselves during the computation process. However, for a very small number of combinations of numbers, the error would persist to the final result.

Intel had decided that the problem was not significant enough to require immediate rectification, for the simple reason that most people never used the FPU enough to notice the problem before the computer would reach the end of its useful life. Perhaps only a few hundred engineers and mathematicians would ever do the sort of intense number crunching affected by the bug. To redo the chip masks and destroy the existing Pentium chips, including chips already shipped, just for this tiny minority of users would not be economical. As a result, Intel decided to correct the bug silently in a planned “step” update of the Pentium.

The decision proved to be a disastrous public relations blunder. In a world that operated on Internet time, unsolved problems were no longer a private matter between customer and company. Six days after his unsatisfactory help desk experience, Nicely e-mailed colleagues around the country to ask for help reproducing the error. Soon the problem appeared on the Usenet newsgroup comp .sys.intel, and from there news of the Pentium glitch expanded exponentially. In an attempt at damage control, Intel offered replacement chips for those whose work required that level of accuracy. However, the computer science community responded to this minimal effort with a combination of scathing humor and outrage.

By the end of November, the story broke to the general public. Ordinary users became outraged that they were considered unworthy of receiving replacement chips. All Intel’s attempts to reassure the common customer that they would never be affected by the bug only seemed to make matters worse. On December 19, 1994, Intel finally announced that it would give a replacement processor to anyone who requested it.

Significance

The Pentium chip was a major technological step forward, containing a superscalar architecture that doubled the microprocessor’s effective speed. However, the fiasco of the Pentium bug was, in many ways, even more significant. It forced Intel to move beyond a purely engineering-based manner of thinking and to recognize the importance of good customer relations. It also had an effect on the computer industry as a whole, ending the old practice of keeping bugs a secret to all but a few large and valued customers. Not only was Intel pressured into publishing a complete list of known bugs in its flagship processor, but also rivals such as Motorola and IBM found it impossible to resist the pressure to do likewise. As a result, the nontechnical public became increasingly aware of the complexity of modern microprocessors and the sheer impossibility of producing one completely free of all bugs.

Subsequent generations of the Pentium chip, including the Pentium Pro and Pentium III, steadily increased in computing power. Later additions to the Pentium family moved away from that name, instead using such name as Xeon, Core, and Duo Core, but they continued to use the X86 instruction set. These chips proved so successful that even longtime holdout Apple began to use Intel chips on models of the Macintosh computer.

Bibliography

Jackson, Tim. Inside Intel: Andy Grove and the Rise of the World’s Most Powerful Chip Company. New York: Dutton, 1997. Solid corporate history. Focuses on the Grove years, including the Pentium.

Reid, T. R. The Chip: How Two Americans Invented the Microchip and Launched a Revolution. New York: Random House, 2001. A basic history of the development of the microchip and the founding of Intel.

Riordan, Michael, and Lillian Hoddeson. Crystal Fire: The Birth of the Information Age. New York: W. W. Norton, 1997. Good background. Helps place the microchip in the larger context of the convergence of information technology.

Seitz, Frederick, and Norman G. Einspruch. Electronic Genie: The Tangled History of Silicon. Urbana: University of Illinois Press, 1998. Includes information on Grove’s role in Intel’s leadership.

Yu, Albert. Creating the Digital Future: The Secrets of Consistent Innovation at Intel. New York: Free Press, 1998. Concerns the development of the Pentium in the context of Intel’s role as industry leader.