The first working integrated circuit on germanium was demonstrated by Jack Kilby at Texas Instruments in 1958. This prototype has a transistor (small left dot) attached to two gold wires and a capacitor (middle black dot). The germanium itself, secured on a glass slide, is divided into three resistors by the tabs at the bottom. By showing that all three types of components could work in the same slice of germanium, Kilby offered a way to improve the performance and lower the cost of electronic devices.
Jean Hoerni, a cofounder of Fairchild Semiconductor, invented the first planar, or flat, transistor in 1959. His novel manufacturing approach was to directly imprint semiconducting and insulating channels onto a ¬silicon wafer. The process left intact a protective layer of silicon dioxide that formed naturally on top of the wafer and prevented contamination. The result was the best-performing transistor of its time.
Robert Noyce—another Fairchild Semiconductor cofounder, who later cofounded Intel—saw a way to use ¬Hoerni’s process to combine multiple electronic components, including transistors, on a single piece of silicon. Announced in 1961, this resistor-transistor logic chip was one of the first commercial integrated circuits. It has four transistors (quadrants in the middle). The white lines are metal traces, which connect the transistors to the two resistors below (horizontal blue bar). The Apollo Guidance Computer used the chip.
In 1974, Intel introduced the 8080. With roughly 5,000 transistors, it was the heart of the Altair personal computer.
Four years later, Intel’s 8086 chip, containing 29,000 transistors, established the x86 architecture that still predominates among today’s chips.
Intel’s 386, released in 1985, had 275,000 transistors and allowed a computer to work on multiple applications at the same time. (Its successor, the 486, was Intel’s first chip with a data cache, which stored a subset of memory onboard for faster processing.)
In 1991, AMD released its own 386 microprocessor, with approximately 200,000 transistors, helping bring competition to the industry.
Motorola’s 68000 microprocessor, introduced in 1979, had 68,000 transistors and powered the Macintosh 128K computer. Chips have several layers; shown here is the layer of wires that link transistors (central blue square). Larger wires around the chip connect it to the surrounding integrated-¬circuit package. The photographer teased out the colors in this chip by adjusting the lighting angle. A chip reflects and diffracts different colors of light depending on the width and spacing of its wires.
The Pentium processor debuted in 1993 and had 3.1 million transistors. It used a technique called branch prediction to forecast upcoming instructions, so it could execute them more quickly. It was also designed with multimedia processing in mind.
In the same year, IBM introduced the PowerPC 601, with more than 2.8 million transistors. Developed jointly with Apple and Motorola, the chip was used in the Apple Power Macs.
In 2000, Intel unveiled the Pentium 4 chip, a completely new design; it had 42 million transistors. The two distinct blocks on the right are part of its cache.
Some of the latest chips improve performance by incorporating more than one “core”—the part of the chip that handles instructions—into a single circuit. Putting multiple cores on a chip is one way manufacturers have been able to increase transistor count without also increasing power consumption. Intel’s 2007 Core 2 Duo has 410 million transistors and a large data cache (big orange block).
The new Core i7 is Intel’s latest four-core chip, with roughly 731 million transistors. It has a shared cache (block along the bottom); its cores, above the cache, make up most of the rest of the chip.
AMD’s Phenom II, scheduled for release in early 2009, has four cores (flanking the rightmost yellow blocks), a large shared cache, and around 758 million transistors.