Before microchips, computers stored bits as sound waves in mercury
Early computers remembered things in some strange ways. One design sent pulses of sound rippling through tubes of mercury; another painted dots of charge on a cathode ray screen. The path from those fragile tricks to today's single-transistor memory cells runs through magnetic rings, a 1967 patent and a chip launched in October 1970.
Memory is where a computer keeps the programs and data it is using right now, as opposed to slower, cheaper and roomier mass storage. In the early 1940s capacity was often just a few bytes. ENIAC, built from thousands of vacuum tubes, could hold 20 numbers of ten decimal digits each. J. Presper Eckert then devised delay-line memory: a mercury-filled glass tube capped with quartz crystals that converted bits into sound waves and back, holding a few thousand bits at most.
In 1946 two rivals arrived that both used electron beams inside glass. Fred Williams's tube, built from a cathode ray tube, became the first random-access memory and held thousands of bits against the Selectron's 256, though it was notoriously sensitive to disturbance. Magnetic-core memory followed, keeping its contents when the power failed. Frederick Viehe and An Wang developed it in the late 1940s, Jay Forrester and Jan Rajchman refined it, and it reached the market in the Whirlwind I computer in 1953, staying dominant into the 1960s.
Chips took over gradually. Texas Instruments shipped semiconductor memory built from discrete parts to the US Air Force in 1961, and IBM introduced its SP95 memory chip in 1965. John Schmidt at Fairchild created MOS memory in 1964, cheaper and less power-hungry than core. Then Robert Dennard at IBM realised a tiny capacitor could store a bit as charge or no charge, with one transistor controlling it. He patented that single-transistor DRAM cell in 1967, leading to Intel's 1103 chip.
Today the split is mostly between volatile and non-volatile types. Static RAM needs six transistors per bit and serves as processor cache; dynamic RAM uses one transistor and one capacitor, must be refreshed constantly, and packs far more bits cheaply into desktop machines. Flash, which Fujio Masuoka invented at Toshiba in the early 1980s, keeps data without power.
Source: Computer memory
