Reconstruction of the Atanasoff-Berry computer
John G. Gustafson · The first computers · 2000
The Atanasoff-Berry Computer (ABC) introduced electronically switched binary logic to computer design in the late 1930s. It was also the first to employ dynamically refreshed capacitors for main storage, the approach used in current dynamic RAM. Perhaps most astonishing is that it was conceived as a parallel computer with 30 simultaneous operations, yet had far fewer parts than the serial computers that followed it in the 1940s. The final version was built in 1939–1942, but was later dismantled. Only a few parts of the original computer remain. In 1994, a team of engineers, scientists, and students began rebuilding the ABC, and were able to demonstrate the functioning replica by October 1997. While patent applications, purchase orders, photographs, and other descriptions gave precise information for the reconstruction effort, there were a number of mysteries about the computer that were solved as a result of the effort to reproduce the machine. The answers to those mysteries are presented here for the first time. We also describe the computer, as much as possible, in modern computer architectural terms to facilitate technology comparison. 1. Atanasoff’s Motivation John Atanasoff was a physicist by training, whose primary interest was crystallography. As a graduate student he often found it necessary to solve systems of linear equations by hand, and as a professor he had advisees similarly engaged in the arduous process. He noted: Since an expert [human] computer takes about eight hours to solve a full set of eight equations in eight unknowns, k is about 1/64. To solve twenty equations in twenty unknowns should thus require 125 hours... The solution of general systems of linear equations with a number of unknowns greater than ten is not often attempted. —Computing Machine for the Solution of large Systems of Linear Algebraic Equations J. Atanasoff, 1940 A strikingly similar observation was made by Alt eight years later [Randell]: ...13 equations, solved as a two-computer problem, require about 8 hours of computing time. The time required for systems of higher order varies approximately as the cube of the order. This puts a practical limitation on the size of systems to be solved... It is believed that this will limit the process used, even if used iteratively, to about 20 or 30 unknowns. —A Bell Telephone Laboratories Computing Machine F. Alt, 1948 In the evolution of automatic computing, solving a system of linear equations was the original “Grand Challenge.” Trigonometric functions and logarithms could be tabled, and mechanical desktop calculators such as the 10-decimal Marchant could handle short sequences of + – × ÷ operations. But Gaussian elimination had order N operations for N unknowns, and this made manual methods unscalable. Atanasoff listed his target applications, which look much like the typical workload at a modern supercomputer center: Multiple correlation Curve fitting Method of least squares Vibration problems including the vibrational Raman effect Electrical circuit analysis Analysis of elastic structures Approximate solution of many problems of elasticity Approximate solution of problems of quantum mechanics Perturbation theories of mechanics, astronomy, and the quantum theory Atanasoff struggled for years to find a physical way to perform arithmetic that was digital instead of analog; he appears to have been the first to draw the distinction between the two. In late 1936 or early 1937, he suddenly made the mental leap that provided him with what he was seeking while at a roadhouse near the Mississippi river. He jotted four design principles [2] on a napkin, paraphrased here: Electricity and electronics, not mechanical methods Binary numbers internally Separate memory made with capacitors, refreshed to maintain 0 or 1 state Direct 0-1 logic operations, not enumeration On the other side of the world, the designs of Konrad Zuse were independently paralleling those of Atanasoff, but the Zuse designs were mechanical or relay-based. By 1940, Atanasoff and graduate student Clifford Berry had taken the above ideas to practice.We now present details of the design, noting features not previously explained in the literature on the ABC.