220 | 21 ACE
Another machine was exported to the New South Wales Institute of Technology in 1956.
Generally, however, DEUCE was not popular in higher education. Computer centres in uni-
versities had two missions: to provide a computing service and to train undergraduates in
programming. DEUCE was excellent for the first mission, but was horrendously difficult to
program for a novice. As a result, a competing machine, the Ferranti Pegasus, was a much more
popular choice for universities. The Pegasus had been specified by Christopher Strachey and,
informed by his early experience with the Pilot ACE, he had avoided confronting the user with
the need for optimum programming. The Pegasus was a drum-based machine with a similar
specification and price to DEUCE, but by a neat architectural innovation Strachey achieved
approximately half the performance of DEUCE without the need for optimum programming.
The Pegasus gained a well-deserved reputation as a programmer’s dream machine.
In economic and industrial terms, the most important legacy of the ACE design was the
Bendix G-15 computer, of which over 400 were sold in the United States. In a parallel develop-
ment to the English Electric DEUCE, the Bendix Aviation Corporation in California wanted
a computer for its own use and as a potential product. In order to specify the machine, Bendix
retained Harry Huskey as a consultant (he was then employed by Wayne State University in
Detroit). The first G-15 was delivered in 1956 and sold for a modest $45,000 (about £16,000).
The machine dispensed with the difficult and expensive mercury delay-line technology and
instead used a magnetic drum as its primary store. The G-15 drew much from the ACE design,
and optimum programming gave it an acceptable speed despite the slow memory technology.
For a decade it was one of the workhorses of the US aerospace and engineering industries.
Although there were other direct derivatives of the ACE design (including the one-of-a-kind
MOSAIC at the TRE and the Packard-Bell 250 in the United States), none was as important as
DEUCE or the Bendix G-15. All of the derivatives of ACE had an exceptional price- performance
ratio, thanks to optimum programming.
In the United States optimum programming was used most prominently in the IBM 650
magnetic drum computer, of which 2000 were sold in the second half of the 1950s. IBM’s presi-
dent called it the company’s ‘Model T’ and it established the firm’s dominance of the computer
industry: it is unlikely, however, that this was directly or indirectly influenced by Turing’s work.
By the mid-1950s optimum programming was a well-known and frequently re-invented tech-
nique used in several machines. Moreover, programming techniques had been developed to
automate optimum coding which, along with high-level languages, made the machines as easy
to use as more conventional designs. Optimum coding was an idea of its time, however, and it
disappeared when true random-access memories came along.
Conclusion
The Pilot ACE, and a DEUCE acquired from English Electric which replaced it in 1956, pro-
vided the main computing resource for the NPL throughout the 1950s. Work on the full-scale
ACE continued, however, and it finally came to life in late 1958. It was a spectacular and elegant
machine—its curved floor plan complemented by the modernist floor-to-ceiling windows of
its surroundings (Fig. 21.2). Sadly, by this date the mercury delay-line memory on which ACE
was premised was obsolete, and with it the raison d’être of the ACE design. Of course, since the
NPL had the machine, it was obliged to use it, which it did until 1967.