The need for a replacement for Atlas was recognised almost as soon as the Science Research Council (SRC) took over the running of the Atlas Computer Laboratory (ACL) from the National Institute for Research into Nuclear Science (NIRNS) in 1965, when SRC was established. At its first meeting, the SRC Atlas Computer Committee (ACC), chaired by Brian Flowers, approved the disc upgrade to Atlas, urged the Laboratory to go to 3-shift working as soon as possible and to come forward to the next meeting with proposals for the future. The Flowers Report was awaiting approval by government and it was recognised that this could have an impact on the future role of the Laboratory.
At its second meeting, the Committee noted the improvements to university computing that would arise from the implementation of the Flowers Report but believed the need for a national facility would not go away. However, competing projects within SRC indicated that money would not be available to purchase a replacement until 1971. The Atlas service should be moved to 7-day working as soon as possible and that it was highly desirable that a major new British machine should be ordered for delivery, if possible by 1971. The emphasis on British was partly because this was the national laboratory but also because there was still a belief that the UK computer industry could be a significant international player in the future. The government was aiming to create a major computer company based around the major players in the currently fragmented UK industry. The merger of ICT and English Electric in 1968 to create ICL meant that a UK computer would have to come from that company at a time when it was trying to define a future for its own 1900 Series and the System 4 (similar to the IBM 360 architecture) range from English Electric.
By 1967, ICL had put forward to the Ministry of Technology a proposal to build a top-end addition to the 1900 Series called the 1908 (initially P51). The aim was that should be competitive with the CDC 7600 and IBM 360/195 coming from the USA. ICL believed this could be achieved by having a multi-processor architecture with each having a power of about 10 Atlas and thus a 2-processor system would have similar power to a CDC 7600. ACC wanted to get started on the replacement programme for Atlas and proposed the purchase of an ICL 1906A which would initially be stand-alone but later front-end an ICL 1908A with the long term objective of a dual processor 1908A. ICL presented the case for the initial 1906A-1908A configuration to the Committee in July 1968 and this was accepted. The first 1906A was under power at ICL and the Committee were convinced that ICL could manufacture the 1908A although there were concerns regarding the operating system for such a system.
Quite a few of the Laboratory staff were unconvinced that this was the right approach. The 1900 architecture was showing its age and staff were unconvinced as to its future. Probably the best expression of these doubts was a letter from Mike Baylis to Jack Howlett in January 1967.
After a year progressing the system, ICL announced to the Committee in July 1969 that it had decided to abandon the 1908A. Peter Hall of ICL said they could not continue to maintain two incompatible ranges (ICT1900 and EE System 4) and after a detailed study had decided to base their new range on the latest Manchester University architecture and they would build a top-end machine initially, comparable in power to the 1908A. By starting near the top, they would avoid the problems of stretching a small-machine architecture upwards. The new machine would be simpler and physically smaller than the 1908A. ICL therefore had decided to concentrate on the new series and abandon the 1908A. Efforts would be made to make transfer of programs from both System 4 and 1900 easy.
The Committee decided to continue with the purchase of the 1906A with the aim of using that as the front-end to the new ICL top-end machine, called initially P.52. This later became the ICL 2900 Series.
ACL continued to look at possible high-powered computers that could be used as a back-end to the ICL 1906A. As well as the ICL New Range, this included the CDC 7600 and their new parallel processing machine, the STAR-100.
In consequence, the ICL 1906A that replaced Atlas was only twice its power but with the possibility of of another system arriving to increase the batch processing power. As an interim solution, ACL was allocated 20% of the computing power of the IBM 360/195 to be installed at the Rutherford Laboratory next door.
