In the mid-1960s, the basic element of computer memory was magnetic core storage, which comprised hundreds of thousands of doughnut-shaped ferrite elements, each about half a millimeter in diameter, laced into rectangular arrays with thin copper wires. Although their assembly had become highly automated, the achievable density, cost and performance of magnetic core systems was reaching a limit — even as the demand for computer memory was skyrocketing.
A practical alternative was emerging in the form of metal-oxide semiconductor (MOS) technology, in which data is stored on a memory chip. Dennard was working on an MOS project that, while promising, was highly complex — each memory cell used six transistors. The system worked, but it required considerable real estate on an integrated circuit to store just 1 bit of data. Worse, it was slow. The goal was to design something faster, more elegant and cheaper.
Dennard found inspiration in a presentation he saw earlier in the day by a group of IBM researchers who were trying to shrink magnetic memory to a compact 25-centimeter square. He followed their lead and sketched out a plan for a system that would hold a bit in a single transistor. It would store bits as positive or negative charges on capacitors — energy-storage devices within a miniaturized electronic circuit — which would be refreshed repeatedly by tiny amounts of energy. The charging process was a key stroke of brilliance. It’s what enabled the capacitor to maintain the data, and it made the RAM system “dynamic.”