lcamtuf has posted a history of working memory on blog.coredump.cx, arguing that RAM gets far less attention than Babbage and that memory constraints held back personal computing until the end of the 20th century.
The post starts with relay latches, including the roughly 2,000 relays in the memory unit of Konrad Zuse’s Z3, against about 600 for the rest of the machine. It moves to the 1918 Eccles-Jordan bistable tube circuit, the ancestor of the SR latch cells in SRAM today.
Tubes and relays were too expensive for bulk memory, so machines used dynamic schemes instead: the spinning capacitor drum of the Atanasoff-Berry computer, mercury delay lines like the one in UNIVAC I, torsion wire memory in the Monroe 925 calculator, and the Williams and RCA Selectron tubes that stored bits as visible dots on phosphor.…