Riding the Whirlwind

Early in 1944, MIT professor Jay Forrester began a project to construct an analog flight simulator. This project represented the confluence of wartime funding opportunities and the desire to explore automated control in the interface between "men and machines." As the nature of the task became clearer with simple experiments, by 1946, Forrester submitted a new proposal to the U.S. Navy's OSRD (Office of Scientific Research and Development) to construct a digital machine which could successfully attack the multiple simultaneous linear equations in many unknowns required of a flight simulator/stability analyzer.

Competed in 1949, Whirlwind is one of the history of computing's landmark machines. Containing 3,330 tubes (one fifth that required in ENIAC, completed just three years before) and 8,900 diodes, Whirlwind first ran for an entire hour of trouble-free operation in June of 1950. Whirlwind used a form of memory popular at the time known as "Williams Tube" memory. Williams Tubes were basically cathode ray tubes (as might be used in an oscilloscope) with a fine mesh covering the surface of the tube. As dots (and "undots") were written to the surface of the tube, the mesh was able to detect the resulting electromagnetic force and feed this back to an amplifier. The system worked by recirculating this bit stream between the CRT and the amplifier/sense circuits.

Williams Tubes were notoriously unreliable and sensitive to environmental conditions, requiring what one early designer called "black magic" to make them work reliably over any period of time. IBM's first commercial computer, the Model 701 (introduced in 1953) for example, used Williams Tubes and even had a special function to "CHECK STOP" the machine every 15-20 minutes to allow a dump (or image) of the current Willams Tube memory contents to be punched into punch cards to reduce data loss should the machine fail. In fact, at the public unveiling of the 701, IBM founder and chairman Thomas Watson Senior posed for photographers in front of the machine. The photographers' flash bulbs saturated the William's Tubes photosensitive surface bringing the machine to a grinding halt. Watson's raised eyebrow—the equivalent of the Evil Eye at IBM—spurred a quick fix: the rapid application of an opaque cover over the tubes.

Whirlwind grew larger in the ensuing years as features (such as larger memory and marginal checking) were added and as Forrester cagily re-aligned his program objectives with what the Navy would support. The most significant development in terms of computing history, however, was the complete replacement of the flaky Williams Tube memory with magnetic core memory in 1953.

Core memory, developed in several places concurrently, but perfected by Forrester at MIT, was made up of small powdered iron oxide donuts (or "cores"). Each core had an X and a Y address line with one (and sometimes two) sense lines running through it as well. Core proved extremely reliable, was relatively easy to manufacture (initially IBM converted an aspirin-making machine to stamp them out) and was inexpensive and non-volatile. This non-volatility makes core a popular choice among spacecraft designers even today where its inherent radiation hardness and ability to retain information even after a power failure make it ideal as a solution to the rigors of space travel.

Whirlwind was also one of the first generation of electronic, general-purpose computing machines and occupied an entire two-story Building (the Barta Building for any MIT alums). A single bit of the ALU took an area two feet wide and twelve feet high so the image of Whirlwind is one of row upon rows of racks-and with the total computational power of less than a Palm Pilot today. The second floor basement was devoted to a drum storage area (like a disk drive but using a cylinder rather than a disk) and Whirlwind's enormous power supplies. Total power consumption was about 150 kW—or enough to power 150 average American homes today.

Whirlwind went on to shape the 700 series of machines at IBM and led to the MIT Memory Test Computer and the transistorized TX-0 and TX-2 machines. Much was learned about designing high-availability (for the time) and reliable machines on Whirlwind. But its more important legacy is its impact on the group of graduate students who worked on it and later machines—students who viewed computing machinery as something one had a personal interaction with (rather than through the more-faceless and paternalistic batch mode of processing typical of IBM systems) and who went onto start Digital Equipment Corporation—birthplace of the minicomputer.

Parts of Whirlwind (including parts of its core memory) are on display at The Computer Museum History Center in Mountain View, California.

Joe Thompson, one of the Whirlwind operators, sits at the computer's console.

Whirlwind's core memory was composed of this matrix of small, powdered iron-oxide doughnuts.