• First real-time computer,
• First computer with a graphic display,
• First computer using core memory,
• First electronic computer not built as a replacement for a
The project that led to the building of the MIT (Massachusetts Institute of Technology) Whirlwind computer was a Navy contract during World War II, aimed at creating a real-time flight simulator with a graphics display simulating the aircraft control panel. Jay Forrester, the project lead, first tried building an analog computer for the task, but concluded that it could not provide the necessary accuracy in the aerodynamics model. Jerry Crawford, who saw the ENIAC demonstrated in 1945, suggested that they build a digital computer instead. The design was completed in 1947, and construction started with a team of 175 people, including 70 technical staff.
Whirlwind went live on April 20, 1951. By this time the Navy had lost interest, since the Whirlwind was not going to be fast enough for their training system, but the Air Force took their place, funding Whirlwind development under Project Claude.
Processor speed was initially 20 KIPS (Thousands of Instructions Per Second), limited primarily by the access time of the Williams-Kilburn CRT storage tube used for main memory. It stored bits in the form of activated phosphor dots on the screen, reading and refreshing them before the dots faded. Each dot could be read due to the charge on the face of the tube. The tubes could be operated on a 4-step cycle, synchronized with the 4-step instruction processing cycle. Except during the Fetch Instruction step, a memory operation and a separate instruction process would be carried out simultaneously
• Refresh a line; increment instruction pointer
• Fetch instruction
• Refresh next line; decode instruction
• Random Access Read/Write; execute instruction
The original MIT Whirlwind filled four floors of a two-story building, counting the megawatt power substation in the basement and the HVAC system on the roof that had to extract all of that heat. The processor was built from 18,000 vacuum tubes drawing about 50 watts each. When the tube racks were fully powered up, it was not safe to walk between them due to the heat they gave off.
Every day, the first task after turning on the power was for someone to check the tube racks and replace any burned-out
tubes. Then a program entered on a bank of toggle switches in binary machine language could be run. The length of time between tube burnouts was the limiting factor on the length of programs that could be run. When external storage was added, the toggle switches were kept in a boot loader program, effectively acting as system ROM.
The most famous program written for the Whirlwind I was the bouncing ball routine, which showed the correct gravitational arc, and loss of energy on each bounce. A straight line represented the ground or the floor. Leaving out a segment of this line where the ball came down would cause the ball to continue falling off the bottom of the screen.
The invention of core memory, which is inherently faster and does not require refreshing, allowed Whirlwind to run at 40 KIPS starting in 1953. This was fast enough for the Air Force SAGE (Semi Automated Ground Environment) air defense system. The contract to manufacture production versions of Whirlwind was initially given to RCA, and then to IBM. System design was done by the MITRE Corporation, an MIT spinoff, and Western Electric’s Air Defense Engineering Service (ADES) got the system integration contract.
Thanks to Edward Cherlin, Simputer Evangelist, for all this information.
Edward's father, George Yale Cherlin, Ph.D., worked on programming the original Whirlwind I as a graduate student in 1951, contributing to the famous Bouncing Ball program.
Picture from www.cedmagic.com