The first stored-program computer was built mainly to test its memory
On 21 June 1948, a machine in Manchester nicknamed the Baby ran the first program ever held in electronic memory. The Baby was not the main goal. Engineers built it as a testbed to prove that their memory, glowing charged dots on the face of a cathode-ray tube, could be trusted.
That device, the Williams tube, was the first practical random-access memory: an electron beam could read or rewrite any spot on the screen in any order. It held only a few hundred to about a thousand bits, but it beat the alternatives. Earlier machines relied on relays, rotating drums whose speed depended on knowing where data physically sat, or ultrasonic delay lines that could only replay bits in the order they went in. Random access means every location takes roughly the same time to reach, unlike a spinning disk or a tape that must wind to the right place.
From the late 1940s, magnetic-core memory took over: grids of tiny magnetised rings, each holding one bit set by the direction of its magnetism and selected by crossing wires. It remained the standard until the early 1970s. Semiconductor memory arrived in the 1960s, first with fast but expensive bipolar transistors. IBM put a 16-bit chip into its System/360 Model 95 in 1965, and Toshiba's BC-1411 calculator stored 180 bits in transistor-and-capacitor cells. Neither could match core on price until MOS transistors, cheaper and less power-hungry, arrived and overtook core in the early 1970s.
The decisive idea came from Robert Dennard at IBM in 1966: a single transistor paired with a tiny capacitor, charged for 1 and empty for 0. That design, dynamic RAM, packed bits far more densely than static RAM, which uses around six transistors per bit. The catch is leakage. Each capacitor's charge drains away, so DRAM must be refreshed every few milliseconds, and everything vanishes when power is cut. Intel's 1103 of 1970, holding 1 kilobit, was the first commercial DRAM chip.
Today the two types split the work. Fast, costly static RAM serves as the processor's cache, while cheaper DRAM forms main memory. Samsung released the first commercial double-data-rate SDRAM in 1998. Because chips are selected through binary address lines, memory sizes almost always come in powers of two.
Source: Random-access memory
