"The Busicom people arrived on a Friday (June 20th 1969). That night I went to
The Busicom team were still there when Hoff got back from
"My job was to act as liaison - not really the design part", says Hoff, "three engineers would wrap up the design and then go to MOS. I was reasonably familiar with computer architecture - though I knew little about calculator design - and it was thought reasonable that I should help out."
Busicom's engineers might have been saved some soul-searching if they'd hadn't waited.When Hoff did return from
"I couldn't believe how complicated the logic for the thing was - pages and pages of schematics" recalls Hoff, "there were so many chips to be done - a separate chip for the keyboard, a separate chip for the display control, a separate chip for printer control, a ROM for programming." In all there were twelve chips in the Busicom design.
At the time Hoff was using a DEC PDP-8 minicomputer for his research.
"I wondered why the calculator's logic should be so much more complicated than the computer's", says Hoff, "as I remember, the ROM had pretty high-level instructions - the equivalent of a floating point add - and it seemed to me that they could simplify the logic if they moved more functions into ROM and made simpler hardware."
Hoff's idea was to put all the processing logic into one chip and attach to it a couple of memory chips. One memory chip held all the data and the other stored the programme to drive the processing logic. A shift register performed the I/O. His approach reduced the chip count from twelve to four.
More significantly Hoff had conceived a general purpose computer that could control anything from a traffic light to an electric motor.
"Now we can make a single chip and sell it for several thousand different applications," said Gordon Moore at the time.
How right he was
NEXT FRIDAY (July 18th): Masatoshi Shima of Busicon tell how 4004 got designed