On Fri, 2024-04-12 at 16:13 -0400, Paul Koning via cctalk wrote: > Not all that fast, well, it depends on what you're comparing with. > Given tube logic with cycle times measures in microseconds, quite > possibly serial rather than parallel organization, those acoustic or > drum memory systems weren't all that terrible.
Speaking of drum machines.... On the Bendix G-15, memory was logically 20 "long" delay lines, each of 108 words of 29 bits, and four "short" lines of four words each, 27 times faster, used for registers. The "delay line" was not an acoustic mercury delay line. Rather, it was implemented on a drum as a digital delay line, with data read and them immediately written a short distance away on the same track. Data were not retained on the drum when the machine was turned off. It was a serial machine. Even the instructions were designed to minimize how much had to be held in flop flops. The machine was remarkably cost effective for its time. Being five feet tall with a one square yard footprint, it could have been a home computer, but as far as I know, the only "home" version was one that the designer, Harry Huskey, got. It was based on the Automatic Computing Engine, or ACE, designed by Alan Turing. Huskey had worked with Turing for about a year. Huskey also worked on SWAC. He did most of the G-15 design work while he was a professor at UC Berkeley. Nicklaus Wirth was one of his grad students. David Evans, cofounder of Evans and Sutherland, worked for him on the G-15 project.