Stephen W. Dunwell led the engineering effort in the development of IBM's STRETCH "supercomputer," which was first shipped in 1962 -- two years before the introduction of the legendary IBM System/360.
The development of STRETCH illustrates the kind of engineering problem IBM has traditionally undertaken in advanced technology applications in large-scale computers. What made the STRETCH so complicated was that many more things than ever before had to go on simultaneously in one computer.
As Dunwell later recalled: "For example, a high speed disk file might be pouring information through a memory bus, a kind of interconnecting highway system, which must simultaneously carry information for the arithmetic unit. We needed the interconnection bus to handle data traffic to and from several memories and other parts of the machine. A device called a program look-ahead controlled this traffic system."
STRETCH's story began in 1954. Discussions in the fall of that year attempted to identify systems problems suggested by computer design and user experience up to that time and to explore the applications of the rapidly developing transistor technology. At first, the planners concerned themselves with small computers but by early 1955, it was realized that the most useful application of the developments necessary to achieve the required translators, circuits, memory and packaging would be in a large rather than a small computer.
STRETCH was thus conceived as a "supercomputer," one that would stretch the performance of the electronic computer far beyond that of the technology then known. It was also to stretch the IBM company itself into new dimensions of research and development and give the pursuit of technology new structure and purpose.
The "supercomputer" became the focal point of research and engineering efforts in transistors, circuits, circuit cards and back panels, memories, frames and power supplies. Later, many of the R&D activities, such as the memory, tape, and circuits groups, were spun off as major development laboratory units.
In October 1970, Dunwell -- then an IBM Fellow -- was honored in a surprise luncheon hosted by Jerrier A. Haddad at the IBM Poughkeepsie Development Laboratory. Attending were members of the original STRETCH team, including John A. Hipp, Robert P. Fletcher, Herbert K. Wild, Werner Buchholz, Erich Bloch and J. Douglas Calvert. Dunwell was presented a plaque describing the STRETCH K1 as "the best computer in the world -- outstanding performance and reliability for eight years."
STRETCH K1, one of six STRETCH systems built, was installed in the Lawrence Radiation Laboratory in Livermore, Calif., in February 1962, for research and development in nuclear and thermonuclear energy. The computer filled six large moving vans and weighed an estimated 180,000 pounds. Each of the six magnetic core memory units for the machine weighed one ton. That machine was still operating in 1970.
Stephen Dunwell's achievements weren't limited just to the development of STRETCH, as significant as that was. In February 1992, after a half-century of secrecy, IBM and Dunwell disclosed that he had helped to develop a top-secret IBM computer that decoded intercepted enemy radio messages during World War II.