IBM 701

• 1. Introduction

  ›

  The IBM 701 Electronic Data Processing Machine arrived in a world that had barely begun to imagine what a computer could do. Announced to the public on the 21st of May 1952, this machine carried a revealing nickname during its development: the Defense Calculator. That name points straight to who needed it and why. But the 701 became far more than a military tool. It translated Russian sentences into English, played a passable game of checkers, and helped researchers figure out the best way to play blackjack. How did a single machine touch all of those corners of human life? And what did it take to build the first commercial scientific computer that IBM ever sold?

• 2. A Machine Born of Need

  ›

  Jerrier Haddad and Nathaniel Rochester designed the 701, drawing on the architecture of the IAS machine at Princeton. IBM had a commercial history rooted in business accounting, not scientific calculation, so this represented a genuine departure. Before a single unit went into production, Thomas Watson Jr. traveled to visit twenty companies that were potential customers. He expected to walk away with orders for five machines. Instead, he returned with orders for eighteen. Watson described exactly that outcome at IBM's 1953 annual stockholders' meeting. It is worth noting what that trip overturned: a comment frequently attributed to his father, Thomas Watson Sr., that there was a world market for maybe five computers. The elder Watson likely never said those words in quite that form, but the younger Watson's own surprise at the demand shows that even IBM's leadership had underestimated the appetite for serious computing power. Nineteen IBM 701 systems were ultimately installed, and the customer list reads like a map of postwar American industry and defense: eight went to aircraft companies, including Lockheed, Douglas, North American Aviation, and Boeing.

• 3. The Defense Calculator at Work

  ›

  At the Lawrence Livermore National Laboratory, the 701 allowed scientists to run nuclear explosives computations faster than had been possible before. The National Security Agency in Washington took delivery of one in 1953, as did the U.S. Navy station at Inyokern, California. The Rand Corporation, General Electric, Convair, and General Motors each put the machine to work on problems that required sustained numerical muscle. In early 1954, a committee of the Joint Chiefs of Staff asked for a direct comparison between the 701 and Remington Rand's competing UNIVAC 1103, specifically to judge which machine should handle a Joint Numerical Weather Prediction project. The two systems turned out to have comparable raw computational speed, with a slight edge to the IBM machine. On input/output, though, the 701 won unanimously: its equipment for moving data in and out was significantly faster. That practical advantage mattered enormously when real work meant reading thousands of punched cards and writing results back out.

• 4. Inside the Machine

  ›

  Seventy-two Williams tubes formed the heart of the 701's memory, each one three inches in diameter. Together they held 2,048 words of 36 bits each, a total built up from 72 tubes carrying 1,024 bits apiece. Williams tube memory was unstable by nature; it required periodic refreshing, which forced the machine to insert refresh cycles into its own timing. A simple addition took five 12-microsecond cycles, two of which were just for refreshing the memory rather than for the arithmetic itself. A multiplication or division demanded 38 cycles, or 456 microseconds. Memory could be expanded to a maximum of 4,096 words by adding a second set of 72 tubes, or later by swapping the whole memory subsystem out for magnetic-core storage. The full system, depending on configuration, weighed about 20,516 pounds. Its instruction set was lean: instructions were 18 bits long, with 5 bits for the opcode allowing 32 distinct instructions, and only two registers that a programmer could access directly. The Magnetic Drum Reader/Recorder was added on the recommendation of John von Neumann, who argued it would reduce the demand for high-speed input/output operations. The first magnetic tape drives on the system had originally been developed for an earlier machine called the Tape Processing Machine before being adapted to fit the 701.

• 5. Experiments That Lingered

  ›

  On the 7th of January 1954, researchers from IBM and Georgetown University ran a program on the 701 that translated sentences from Russian into English. This Georgetown-IBM experiment was the first public demonstration that a computer could handle natural language translation, even in a limited way. The same year, a group of scientists ran millions of simulated blackjack hands on the 701, testing every possible combination of cards to identify the statistically correct decisions for hitting, standing, doubling down, or splitting. The rules they derived remain the standard basis for optimal blackjack strategy. Then on the 24th of February 1956, Arthur Samuel demonstrated a checkers-playing program on the 701 over live television. The program was notable enough that it is still cited as one of the earliest demonstrations of what would later be called artificial intelligence. A self-proclaimed checkers master named Robert Nealey played a later version of the program on an IBM 7094 in 1962, and the computer won. John Backus also used the 701 as the platform for Speedcode in 1953, described as the first high-level programming language created for an IBM computer; Speedcode was built to handle floating-point arithmetic, a capability the machine otherwise lacked at the instruction level. Backus would later develop the Fortran compiler, but that waited for the 701's successor.

• 6. What Came Next

  ›

  The IBM 701 was the first entry in the IBM 700/7000 series, the company's high-end computer line for more than a decade. Its business-oriented sibling was the IBM 702. A lower-cost sibling aimed at general use, the IBM 650, went on to earn a reputation as the first mass-produced computer. The 701's direct successor was the IBM 704, released four years after the 701. The 704 was not backward compatible: it expanded instructions from 18 bits to 36 bits to accommodate new features, which meant programs written for the 701 had to be rewritten. The 704 also marked the transition from Williams tube memory to magnetic-core memory as the standard. The broader 700/7000 family continued as IBM's flagship high-end line until the IBM System/360 arrived in 1964. The University of California Radiation Laboratory at Livermore, one of the 701's most demanding users, built a language compilation and runtime system for their machine called the KOMPILER, an early sign of how the community around the 701 kept pushing the hardware toward new territory.