In the early to mid '60s Don Hitt of IBM's Systems Development Division, a field engineer turned mainframe designer and my first mentor, built in his Poughkeepsie basement exactly the computer who's architecture you suggest. It used a magnetic drum memory and a glass acoustic delay line from one of IBM's office products.
[begin reminiscence]A couple of years later Don and Bob Wasserman designed and built the first microprocessor so far as I can tell: a 4-bit ECL computer with
4K ram. While it was not a a single chip device it had all of the elements necessary to call it a microprocessor. It was also the first and most reduced RISC I've ever heard of. It had four instructions and no arithmetic logic unit, just an accumulator it could load, store and conditionally branch on being zero. The fourth instruction was a branch and link. Using table lookup, storing into the instruction stream and subroutines for ALU operations it was programmed first as a scientific desk calculator driving an oscilloscope as a decimal digit display device.It was a between-project bit of fun for these guys but with serious intent. They incorporated it in our next project, the System/370 Model 158, as its minimal-core maintenance processor to diagnose hardware failures in the mainframe and tell field engineers what logic card to replace. I did a large part of that programming.
It was also used as the first in-house floppy drive writer. The first field use of a floppy was for storing the diagnostic programs that the little processor executed in diagnosing the mainframe. As a floppy writer it was channel attached to a System/360.
You wouldn't believe the number and variety of mechanical designs tried out before the final 8" oxide coated mylar disk in an envelope design that went to the field as a floppy disk.
[end reminiscence]Thanks for triggering that memory. :-)
Bob