 |  |  |  |  |  |  |  |  |  |  |  |  |  |
Intel: defined the microprocessorby Alexander Wolfe From incremental advances in semiconductors, significant technologies have grown.That was certainly the case in the 1950s, little more than a decade after physicists John Bardeen, Walter Brattain and William Schockley of Bell Telephone Laboratories developed the point-contact transistor. That device, for which the three scientists would receive the Nobel Prize in physics in 1956, is considered to be the first solid-state amplifier. As the Eisenhower years progressed, however, engineers had become increasingly adept at building more sophisticated circuitry that combined multiple transistors, capacitors and resistors on a single breadboard. Indeed, the desire to pack more silicon into ever-smaller spaces (sound familiar?) was about to spark two independent inventors to take the technology to the next step. Driving the demand for a denser means of packing electronics componentry was the first generation of fully solid-state digital computers, which contained more than 25,000 transistors and diodes and were so huge that they required large rooms-dubbed "glass houses"-for proper operation and ventilation. The obvious solution seemed to be to put more than one transistor within the packaging for each device, thus reducing the amount of parts that had to be assembled to create a full-blown computer, radar or other large-scale electronics system. Enter Robert Noyce, then of Fairchild Semiconductor, and Jack St. Clair Kilby, of Texas Instruments. Both men independently conceived the integrated circuit (IC) in 1959. More than just reducing parts counts, the IC solved the increasingly costly problem of fashioning the interconnects between discrete transistors. Kilby fashioned his first IC during a company-wide vacation at TI in July 1958. (As a new employee, he wasn't entitled to the time off.) According to a 1980 historical account in Electronics magazine, Kilby "knew it was feasible to fabricate resistors, capacitors and transistors from semiconductor materials . . . His next step was to see if p-n junctions could be fabricated from a single piece of silicon." The account noted that Kilby's bosses were enthusiastic but were skeptical that a working prototype could be made. Kilby, however, took a small-signal germanium mesa wafer with 25 transistors on it, cut it up and created a distributed RC network to make an IC that functioned as a phase-shift oscillator. His next project was to design a flip-flop. Meanwhile, at Fairchild Semiconductor in Mountain View, Calif., in January 1959, research and development manager Robert N. Noyce "entered into his notebook ideas about incorporating diffused or deposited resistors, about isolating devices from one another by means of revers-biased pn junctions and about interconnecting elements through holes in the oxides," according to the Electronics magazine account. About a month later, when Noyce heard about Kilby's work at TI, the race was on to build and improve the IC. Fast-forward to 1967, when an economic slump sparked a mass exodus of employees from Fairchild. Two of them — Noyce and Gordon Moore — formed Intel Corp. They were soon joined by Andrew Grove, a doctoral-level chemical engineer who had emigrated to the United States from Hungary in 1956. (Kilby continued working at TI, although today he is largely retired. Noyce, who became head of Intel Corp., died of a heart attack in 1989.)
The first microprocessor didn't appear officially until 1971. That chip was the 4-bit 4004, from Intel Corp. Interestingly, the 4004 wasn't designed by Intel just to prove it could be done: Rather, in 1969 a customer called Busicom Corp. of Japan, which designed calculators, had asked Intel for a chip set. Intel set to work under the leadership of Marcian "Ted" Hoff, the company's manager of applications research. The 4004 it- self, which was actually a five-chip set, was designed by Intel engineer Federico Faggin. The 4004 saw only limited service. Intel had already begun work on an 8-bit chip. That microprocessor, dubbed the 8008, was released in early 1972. By that time, competitors like TI, Fairchild and National Semiconductor were also designing chips. Motorola weighed in with its first device-the 6800-in 1974. By the mid-1970s, dozens of microprocessors crowded the market. The microprocessor battle sparked a simultaneous race to reduce the size of computers themselves. On the corporate level, minicomputers became faster and more flexible. More radical, however, was the debut of the first microcomputers, which were microprocessor-based machines. Although Intel was actually shipping microcomputer kits in the spring of 1983, the concept didn't really catch fire until the famous "Home-Brew Computing Club" era began in the late 1970s. That coterie of microcomputer enthusiasts was historically responsible for having pointed (though not necessarily directly) the technology toward Apple's Macintosh and IBM's PC. A magazine called Popular Electronics launched a firestorm of activity when it ran a story in the late 1970s on how to build one's own microcomputer from a microprocessor-based bit. That early microprocessor era was competitive in a much different way than the later microprocessor period. The early years were marked more by an air of spirited engineering competition, as rivals attempted proof-of-concept designs. The true maturation of the microprocessor can be traced to the introduction in 1983 of 32-bit processors, the most complex computing devices ever to be fabricated up until then. The most successful of those was arguably Intel's 80386, which debuted in 1986. The mid-1980s also saw the rise of reduced instruction set computing, a technique conceived in 1975 at IBM by John Cocke. RISC was intended to reduce the stress placed on a microprocessor's compiler by streamlining the chip's instruction set. In the real world, however, RISC tended to become whatever a chip designer said it was. That is to say, many RISC designs seemed to be more bloated than the CISC (complex-instruction-set computing) designs they were intended to replace. While Intel's Pentium family dominated much of the 1990s, another trend was coming into vogue: very long-instruction-word (VLIW) computing. In the marketplace, the "purer" VLIW ideas tended to find their way into digital signal processors. VLIW had actually been attempted in the mid-1980s, with little success. Intel revived the concept for its new architecture, dubbed IA-64 and shown in prototype form in late 1999. (Hewlett-Packard Co. collaborated with Intel on the design of the basic IA-64 architecture.) However, Intel pointedly steered clear of the VLIW moniker. Rather, the company dubbed the processor Epic, for explicitly parallel instruction computing. As the new millennium begins, the computer industry is anxiously awaiting the introduction of full-blown Epic servers to see how the complex concept works in reality. The Century of the Engineer: Companies that made a difference
|
| ||||||||||||||||||
