The
IBM Stretch....
Steve Dunwell
the the lead architect of this supercomputer starting back in 1956, and was the man
who saved and restored the Bardavon Opera House in Poughkeepsie, with his
wife Julia... It had been slated to become a parking lot. He
let myself and others put on a magic show there around 78 or so... In later
(retired) years he operated the Dutchess County Multiple Listing service
through his company, Data Center Computer
Services, across from IBM Poughkeepsie, where he was still using a large IBM
system (not sure what model) around 1987. He had me do some computer
work for him around that time related to the PC. A few years later, I
was involved with a local group called the PCCA (professional computer
consultants association) and Steve and his wife Julia were also members of
that and came to the meetings. My impression of him was that he was a
genius. In 1992 he and IBM revealed his efforts in developing a
secret computer during world war II that decoded intercepted enemy radio
messages. (see
page two of the IBM history link)
This view of the STRETCH (IBM 7030 Data
Processing System) probably shows the machine in 1960 or early 1961 prior to
its shipment from IBM's development facility in Poughkeepsie, N.Y., to the
Los Alamos Scientific Laboratory in New Mexico. The Lab, operated by the
University of California for the U.S. Atomic Energy Commission, acquired the
computer for research in nuclear and thermonuclear energy. The operator's
console (foreground) was used to control and monitor the system, the central
processing unit (foreground) was the computer's heart, and the Exchange
(left) provided a new level of efficiency in handling input/output devices
-- and was able to control a peak flow of 800,000 characters of information a
second.
The Stretch design had its roots in 1954
from initial studies in "advanced concepts" by Stephen Dunwell and Werner
Buchholz [Bashe, et al., 1986], and from Nat Rochester's encouragement of
Gene Amdahl to design a new high-performance scientific computer after his
work on the 704 [Norberg Interview with Amdahl, 1986/1989]. The project
started formally after IBM lost an April 1955 bid on a high-performance
decimal computer system for the University of California Radiation
Laboratory (Livermore Lab). Univac, IBM's competitor and the dominant
computer manufacturer at the time, had won the contract to build the
2-megacycle Livermore Automatic Research Computer (LARC) by promising
delivery of the requested machine in 29 months. IBM's bid was based on a
renegotiation clause for a machine that was four to five times faster than
requested, cost $3.5M rather than the requested $2.5M, and proposed delivery
in 42 months. In September 1955, IBM proposed a binary computer of "speed at
least 100 times greater than that of existing machines" to the Los Alamos
Scientific Laboratory and received formal approval of a $4.3M contract in
November 1956 for what would become the Stretch computer. Delivery was
slated for 1960.
Gene Amdahl and Stephen Dunwell were major
contributors to the proposed design; but, when Dunwell was chosen at the end
of 1955 to head the Stretch project with Amdahl assigned a lesser role,
Amdahl chose to leave the company. Dunwell recruited Fred Brooks, John Cocke,
and Jim Pomerene in the summer of 1956 to join the project, and Harwood
Kolsky from LASL joined the team in the summer of 1957. Robert Blosk and
Gerrit Blaauw joined IBM in 1953 and 1955, respectively; both joined the
Stretch team in 1957.
More:
http://www-03.ibm.com/ibm/history/exhibits/builders/builders_dunwell2.html
Selected reviews of notebook and powerful PCs, along with instructions on
building your own powerful PC.
