Transistorized Computer
A transistorized computer is a digital computer whose logic circuits are built from discrete transistors rather than vacuum tubes, the defining technology of the so-called second generation of computers, which ran from the early experimental machines of 1953 to the mid-1960s. Each logic gate, flip-flop and amplifier is assembled from individual transistors, diodes, resistors and capacitors soldered onto circuit cards, and thousands of such cards are wired together into a processor; the approach sits between the tube machines it replaced, such as ENIAC, and the integrated-circuit machines that followed.
The first transistor computers were experimental, and several arrived within a few years of one another. The University of Manchester demonstrated a prototype transistor computer in 1953, and Bell Labs built TRADIC for the US Air Force in 1954, a machine that was later flown in aircraft. In the United States, the TX-0 at MIT, finished in 1956, used a large number of Philco surface-barrier transistors and became a favourite of early interactive computing. In Britain, the Harwell CADET followed in 1955. These machines proved the idea and trained a generation of engineers in a technology in which, for the first time, a computer could be small, cool and reliable enough to be considered for aircraft, ships and ordinary laboratories.
Commercial machines followed from the late 1950s, and they made the second generation a market. IBM's 608 calculator, announced in 1955 and shipped from 1957, was the company's first commercial all-transistor calculating machine. The IBM 7090, introduced in 1959 as a transistorised successor to the vacuum-tube 709, became a standard for large scientific and military computing, used for work such as space-mission calculations and ballistic-missile warning. Control Data's CDC 1604, designed by Seymour Cray and delivered from 1960, was a fast scientific machine, and the Ferranti-Manchester Atlas, brought into service in 1962, introduced the virtual memory that later became universal. Smaller machines such as the IBM 1401 and the DEC PDP-1 made transistor computing affordable to businesses and universities.
The advantages over tubes were reliability, size and heat, not just speed. A tube computer needed thousands of filaments glowing at all times, consumed large amounts of power, and failed on a time scale of hours or days because tubes burned out; maintenance staff walked the racks replacing them. A transistor has no filament, needs a fraction of the power, and, if properly designed, can last for years, so the machines could be far smaller and run continuously, with fewer faults and lower cooling cost. Early germanium transistors were not perfect: they were sensitive to temperature and limited in speed, and silicon devices and better diffusion techniques were needed to improve both. Still, the gains in dependability were what made computers usable for business, space and defence.
The limit of the approach was wiring, which pushed the industry toward integration. A second-generation computer contained tens of thousands of individual transistors and diodes and a far greater number of soldered joints, and every one was a place where a connection could fail and a worker had to be paid. Jack Morton of Bell Labs described this problem as the "tyranny of numbers" in 1958: the more components a system had, the more interconnections it needed and the less reliable and the more expensive it became. The response, from Kilby's and Noyce's integrated circuits of 1958 to 1961, was to fabricate the components and their wiring together on a single piece of semiconductor, and the IBM System/360 of 1964 marked the transition to a new generation.
A common misconception is that "second generation" meant a clean, quick break from tubes. Many early transistor machines were hybrids with valves still doing some jobs, and tube and transistor machines were sold side by side for years. It is also easy to credit a single machine with being "the first", when the answer depends on the definition, whether purely transistor, flying, commercial or stored-program. Today the lesson is that every processor, with billions of transistors printed rather than soldered, solves the same problem of reliable, dense, low-power switching, and the discrete-transistor computer survives mainly in museums and in the logic design traditions that those engineers established.
Read transistorized computer as the bridge generation: machines that proved transistors could replace tubes in computing and, by running into the cost and unreliability of wiring them together by hand, made the integrated circuit a necessity.