The Microprocessor Revolution

Abstract

The microprocessor put an entire Central Processing Unit onto a single silicon chip, starting with the Intel 4004 in 1971. What followed was five decades of widening word sizes, from 4 bits to 64, falling prices, and processors turning up in places no one had built computers before: calculators, cars, appliances, telephones, and eventually every device with a battery.

The Integration Breakthrough

Before the microprocessor, a computer’s CPU was a complex assembly of numerous discrete integrated circuits (ICs), transistors, and other components spread across multiple circuit boards. This made computers large, expensive, and difficult to redesign.

In 1971, Intel released the 4004, the world’s first commercially available microprocessor. It integrated all the functions of a 4-bit CPU, the arithmetic logic unit (ALU), control unit, and registers, onto a single piece of silicon.

The Significance of 4-bit Architecture

While modern processors are 64-bit, the 4-bit architecture of the Intel 4004 was perfectly suited for its original purpose: controlling calculators. A 4-bit word could represent a single decimal digit (0-9) using Binary Coded Decimal (BCD), allowing the chip to handle numeric input and output with minimal complexity. This efficiency proved that specialized, small-scale computing could be mass-produced on a single chip.

The chip began as a calculator order. In 1969 the Japanese calculator maker Nippon Calculating Machine (Busicom) sent Masatoshi Shima to Intel, where Ted Hoff and Stan Mazor proposed replacing his many custom circuits with a small general-purpose processor; the contract was signed in February 1970. When Shima returned in April he found that almost nothing had been done until Federico Faggin arrived to design the four chips of the MCS-4 set (see Federico Faggin and the Microprocessor). The 4004 ran in January 1971, entered production in August, went into the Busicom 141-PF calculator in October, and was advertised in Electronic News on 15 November 1971 under the headline “Announcing a new era of integrated electronics.” In August 1971 Intel also launched Dov Frohman’s EPROM, the 1701, a read-only memory that could be erased under ultraviolet light through a quartz window and reprogrammed, which made developing firmware for the new processors practical.

Intel was not alone. Texas Instruments, working for the same customer as Intel’s 8008 (Computer Terminal Corporation, maker of the Datapoint 2200), built the 8-bit TMX1795 under Gary Boone and described it in an advertisement in Electronics on 7 June 1971; it worked, but CTC did not buy it and it never went into production. TI patented the idea and enforced the patents until it lost against Dell in 1992. The independent inventor Gilbert Hyatt filed a patent on 28 December 1970 that was granted in 1990 and earned him licence fees until a TI challenge led the patent office to strip it from him in 1996; who invented the microprocessor remains a matter of definition.

This “computer on a chip” changed everything.

Key Impacts of the Revolution

The ability to pack an entire processor into a tiny footprint had profound consequences for technology and society. This movement toward commodity hardware stood in stark contrast to the era of The Lisp Machine Era, where computing power was often tied to expensive, specialized hardware designed for specific languages. While Lisp machines sought peak efficiency through specialization, microprocessors achieved dominance through general-purpose flexibility and mass production.

Extreme Miniaturization: The microprocessor allowed computing power to shrink from room-sized mainframes to handheld devices. This led to the birth of embedded computing in cars, appliances, and industrial machinery.
Massive Cost Reduction: Producing CPUs on silicon wafers enabled economies of scale. Microprocessors became incredibly cheap, making personal computers economically viable for households and small businesses.
Architectural Versatility: The programmable nature of the microprocessor meant that a single chip could be used in vastly different applications simply by changing the software, driving the rapid diversification of digital technology.

From 4 Bits to 64: The Widening Word

The Intel 4004 was a proof of concept. What followed was a four-decade march through successive generations, each doubling the width of the data the processor could manipulate in a single operation, and with each doubling, entire new categories of software became possible.

The 8-Bit Era: Computers Leave the Laboratory (1972–1980)

Intel’s 8008 (1972) and its successor the 8080 (1974; designed by Faggin and Shima, described in Electronics on 18 April 1974, with more than 4,500 transistors at 2 MHz and a price of $360) extended the 4-bit concept to 8 bits, enabling processors to handle a full byte in one instruction. Motorola’s 8-bit 6800, described in the same issue of Electronics and on sale in November 1974 at the same $360, missed the Altair but went into terminals, cash registers and pinball machines; its price fell to $35 and Motorola had sold 25,000 by September 1976. Several of its designers, Chuck Peddle among them, left for MOS Technology and built the cheaper 6502. Faggin left Intel on 31 October 1974 and founded Zilog with Ralph Ungermann in December; their Z80, financed by Exxon Enterprises, appeared in 1976 as a much-improved 8080. The competitive Zilog Z80 (1976) and MOS Technology 6502 (1975) offered similar capability at low cost. These chips powered the first wave of personal computers: the Apple II (1977, 6502), the TRS-80 (1977, Z80), and the Commodore PET (1977, 6502). The 8-bit era also produced CP/M, the first widely used personal computer operating system, a software ecosystem that briefly seemed destined to standardize the industry before IBM redrew the map.

The 16-Bit Era: The IBM PC and the Architecture That Refused to Die (1978–1985)

Intel’s 8086 (1978) and its lower-cost variant the 8088 (1979) brought 16-bit processing to the PC market. The 8086, started in May 1976 under Stephen Morse as a stopgap while a 32-bit project lagged, was released on 8 June 1978 with 29,000 transistors and a one-megabyte address space; the 8088, with an 8-bit external bus, came from Intel’s laboratory in Haifa in June 1979. Through cross-licensing agreements from 1976 the second-source maker AMD, founded by eight Fairchild employees under Jerry Sanders on 1 May 1969, built the 8088 and 80286; when Intel refused it the 80386, AMD won the right in court in 1991 and sold a million Am386 chips within a year. When IBM chose the 8088 for its IBM PC (1981), it locked in the x86 instruction set as the dominant personal computer architecture, a decision whose consequences are still felt today in every Intel and AMD processor sold.

Simultaneously, Motorola’s 68000 (1979), a cleaner, wider design, powered the machines that many engineers considered technically superior: the Apple Macintosh (1984), the Amiga, and the Atari ST. The 68000 family lost the market battle but influenced language design and operating system architecture for a generation.

The CISC/RISC Fork

As processors grew more complex, a fundamental architectural debate divided the field. CISC (Complex Instruction Set Computing), exemplified by the x86, accumulated hundreds of specialized instructions, reasoning that hardware should do more work per instruction. RISC (Reduced Instruction Set Computing), pioneered by John Cocke at IBM and developed into commercial products by MIPS, SPARC, and ARM, took the opposite view: a small set of simple, fast instructions, executed in a single clock cycle, would outperform complex hardware by allowing faster clock speeds and deeper pipelining. RISC processors dominated workstations and servers through the 1990s; today’s x86 chips internally translate their complex instructions into RISC-like micro-operations, the CISC/RISC debate was resolved by doing both simultaneously. The extreme CISC position had its own casualty: see Dead End: the Intel iAPX 432 and RISC vs. CISC.

The 32-Bit Era: Protected Memory and Multitasking (1985–2003)

Intel’s 80386 (1985) brought full 32-bit addressing, which meant programs could directly access up to 4 gigabytes of memory, and extended the 80286’s hardware-enforced protected mode to 32 bits with paging, isolating processes from each other so that one crashed program could not take down the entire machine. This capability made serious multitasking operating systems possible. Windows NT (1993) and Linux (1991) were both built from the ground up around the 32-bit protected-mode model.

The 1990s saw the architecture wars intensify. Intel’s Pentium (released 22 March 1993 and shown at CeBIT the next day; 3.1 million transistors, 60 or 66 MHz, named so that competitors could not copy a number) introduced superscalar execution (multiple instructions executed simultaneously on separate pipelines) and branch prediction, anticipating which path a conditional instruction would take before it was resolved. Clock speeds climbed from the 8080’s 2 MHz to the Pentium 4’s 3+ GHz by the early 2000s.

The 64-Bit Era and the Multi-Core Transition (2003–present)

The 32-bit address space (4 GB of addressable memory) became a ceiling as RAM grew cheap and applications grew hungry. The transition to 64 bits arrived not from Intel but from its competitor: AMD’s Opteron and Athlon 64 (2003) introduced the x86-64 extension, allowing existing 32-bit software to run unchanged while opening a vastly larger address space. Intel, initially committed to a separate 64-bit architecture (Itanium), adopted AMD’s extension as EM64T in 2004. The industry had, for the first time, followed the competitor’s design.

Clock speed scaling hit a physical wall around 2004, the “Power Wall”: faster clocks generated heat that could not be dissipated reliably. The industry’s response was multi-core processors: instead of one faster processor, pack multiple complete processor cores on a single chip. Intel’s Core 2 Duo (2006) marked the mainstream arrival of dual-core CPUs; server processors now routinely carry 32, 64, or more cores.

ARM and the Mobile Inversion

The dominant architecture of the smartphone era is not x86. ARM (originally Acorn RISC Machine, later Advanced RISC Machines) had spent the 1990s in embedded systems, cheap, power-efficient, and invisible. When Apple needed a processor for the original iPhone (2007), ARM’s power efficiency made it the only viable choice. Within a decade, ARM-based chips powered the majority of the world’s computing devices by unit count.

In 2020, Apple completed the circle by releasing the M1, an ARM-based processor designed in-house that outperformed contemporary x86 laptops in both performance and power efficiency. Four decades of x86 dominance in personal computing began, credibly for the first time, to look uncertain.

The open-source RISC-V instruction set (Berkeley, 2010) adds a further dimension: a royalty-free architecture that anyone can implement, it has gained adoption in embedded systems, custom hardware accelerators, and research, a potential second unbundling of processor design from any single commercial interest.

The Foundation of Modernity

The microprocessor revolution is the engine behind Moore’s Law. Every advancement in semiconductor fabrication has allowed for more transistors on these chips, leading to the exponential growth in computing performance we have witnessed over the decades. Today, from the tiny processor in a microwave to the massive clusters powering the cloud, the legacy of that first 4-bit chip is everywhere.