Electrical engineer and computer pioneer (1919–1995) who designed the first general-purpose electronic digital computer (ENIAC) with John Mauchly, and designed the first commercial computer in the United States (the UNIVAC). J. Presper Eckert's handwritten rough draft and hand-corrected drafts of a speech delivered in Zurich, Helsinki, and Tokyo, circa 1970, highlighted by a 26-page autograph draft (headed "Computers of the 70s by J. Presper Eckert") and a hand-corrected typed draft (entitled "Yesterday, Today and Tomorrow in Computer Technology"), comprising 12 typed pages plus 16 pages of handwritten insertions. Additionally includes a 12-page photocopy of the speech, an 8-page photocopy of galley proofs of the speech, and several pages of original and mimeographed correspondence, telegrams, and papers related to Eckert's travels and speaking engagements.
Eckert's address amounts to a history of computer memory technology since his co-creation of ENIAC, with important predictions for the future. In part: "When John Mauchly and I started in 1943 to study electronic computers, we recognized many new problems. Let me review some of these problems and see where we stand on them today. Before doing that, however, in 1943, a mechanical machine called the Mark I had already been started at Harvard University. This machine was already over 100 times faster than an unaided man in doing arithmetic, but it was electro-mechanical. It could have led to relay machines 100 times faster than itself or 10,000 times faster than man. However, Mauchly and I decided to leapfrog such developments and build a self-operating machine to be called ENIAC. It was 1,000 times the speed of Mark I or 100,000 times the speed of man. Since then, machines have been announced which are over 1,000 times the speed of ENIAC and are thus 100,000,000 times the speed of a human being with pencil and paper.
Perhaps the major problem of building an electronic computer in the beginning was the high-speed memory required. In ENIAC it consisted of tubes, relays, plugs, switches, and resistor arrays. It cost between $25 and $500 per digit depending on whether it was read-only memory, set by switches, or fully electronic flip-flop memory. Today I believe our best solution to this problem, except for small memories, say less than 100,000 bits, is some sort of magnetic thin film memory. The most practical so far is the plated wire memory now in use for several years. This now costs no more than equivalent core memory and is faster, more compact, and takes less power. Present wire memories are hundreds of times faster than most of the memory in ENIAC.
Further development on this wire memory will drive the cost of wire memory down to less than one cent per bit, in sizes of 10,000,000 bits or more. Wire film memories of the 70's will probably be less expensive than core memories and less expensive than semi-conductor memories of equivalent size and speed. We already have experimental wire memory arrays which can read and write in less than one-tenth of a microsecond. Calculations indicate that a memory cycle of a microsecond (50 nanoseconds) seems possible based on present wire and semi-conductor limitations. By interlacing several wire memories in time and by properly anticipating the conditional transfers which occur in a routine, a speed of 5 to 10 times the fastest computers announced today can probably be achieved. And without resort to complex buffer or cache memory arrangements used in some of today’s newest computers." In overall fine condition.
