Apollo Guidance Computer Block II with DSKY

IBM VS. MIT: A Computer Science Showdown

Engineer, Hugh Blair-Smith
Dr. Battin For Brochure
Eldon C. Hall Brochure Photo

The great space grudge match: IBM vs. MIT

Ever watched two computer engineers fight one another? May the sharpest calculator win! In the 1960s the grudge match between two heavy hitting computer engineering teams was at an all time high. On one side of the ‘ring’ was IBM. They were building the code to launch the rocket on the Apollo Missions from Earth to space, called the Launch Vehicle Digital Computer (LVDC). On the other side was Doc Draper and his lab, who had been awarded the contract for the Apollo Guidance Computer (AGC) to navigate the Lunar Module and Command Service Module with the astronaut crew inside. And the judge was none other than NASA. IBM was frustrated that they weren’t in charge of all of the navigation equipment on the Apollo program and commissioned an independent study performed by Bellcom to convince NASA of their superiority. Eldon Hall and Dick Battin (from team Draper) weren’t about to let them win. 

Here is how retired Draper engineer Hugh Blair-Smith describes the battle in his own words from his book Left Brains for the Right Stuff:

"To nobody’s surprise, the study solidly took IBM’s side (having been commissioned by them), concluding that the AGC’s reliability could never match that of the LVDC. Through their proxy, IBM was mounting an invasion of a sacred part of our space, and we’d have to put on our “Churchill face” to repel it. Eldon Hall felt we knew how to defend the reliability point, having spent considerable time with Houston on its two parts. It was clear to me that Bellcomm  hadn’t understood much about how and why the two computers greatly differed in architecture. We had designed the AGC4 to be flexible and efficient in a priority-driven multitasking environment, interacting with a variety of spacecraft subsystems as well as the crew and the ground. IBM had optimized the LVDC architecture for the major continuing function in the Saturn booster’s flight, a guidance method called “adaptive polynomials.” I made up a nasty nickname for the LVDC, “APE,” for Adaptive Polynomial Engine—in parody of Charles Babbage’s “Analytical Engine” of a century before. After coding up some comparative programs in the systems software area, even I was surprised at how much less memory and time the AGC needed.

Dick Battin’s guidance experts performed a similar exercise for powered flight, finding that the AGC produced more precision and better interrupt response and input/output handling. It was about ten times faster than the LVDC while requiring less than half the memory space. Dick, as the designated author of our response, could scarcely believe the strength of our findings. Excellent worry-wart that he was, he kept asking, “Is it really that one-sided?” Everybody on the team assured him, “Yes, Dick, write it. It’s real!”

We put together a lengthy memo and slide show to take to the showdown meeting. . . The NASA folks were eager to see what we’d done, having been given some idea that it would be “interesting.” So after dinner, they came over to the motel to have a preliminary meeting in one of the rooms. Something over a dozen people gathered around two king-size beds, and we spread out our slides and started reading the introduction of our memo. After a very brief outline, our key phrase was, “We are astonished that Bellcomm could have come to this conclusion.”

There were a few moments of thunderous silence, followed by “You are what??” The senior NASA official present said something like, “OK, guys, we’d better hear what this is about.” So we walked them through the slides and our thoroughly one-sided measurements, and the Bellcomm people didn’t have much in the way of artillery to oppose them. A more “official” meeting may have been held the next day in the NASA offices, or not. I don’t remember and it surely didn’t matter."

In the end the two teams were left to each retreat to their own territories. MIT's AGC computer was proved superior in terms of its reliability, memory and function. IBM's LVDC computer would focus solely on supporting the Saturn V booster guidance and not the larger mission and spacecraft.