(By Virgilio Pasquali –November 2003)
The brilliant simplicity of the original FP6000 design and the design skills of the ICT engineers made it relatively easy to expand the design into a series of systems spanning a wide range of performance and price. Unsurprisingly, the greater difficulty was encountered in expanding the range upwards.
The 1906/7, the top model of the original series, had a performance that was significantly below the upper models of the 360 range, and this led to the early introduction of dual processors.
Anonymous dual processors (1906/7E/F) shared the same IAS (core store) via one or more SMAC’s (Store MultiAccess Control) and ran two (or more) programs concurrently without them being aware of each other, under the control of a single operating system. Given a suitable workload, a dual processor could reach almost double the performance of a single.
The multiprocessor design was extended in the 1908 (a four processors system), unveiled at the Edinburgh IFIP in 1968, but overtaken by events and not put into development.
The perceived need to have more powerful single processors to compete more effectively in an area of the market where new advanced applications (such as large real time transaction processing) were emerging, was one of the many factors leading, eventually, to the top down introduction of the 2900 Range in 1974
At the lower end of the range, immediately after the 1901, a system below 1901 was seriously investigated. Initial feasibility studies proved that it was technically feasible to design and manufacture a 1900 compatible system at the right cost and suitable specification. But, given the selling methods used at the time, the selling costs made such a system unprofitable, and the project was dropped.
The viable span of the range (1901-1906/7) was thus established.
The following chart plots the commercial performance of each (single) processor in the ICT 1900 and IBM 360 in1966, measured by instruction mixes (POWU2/sec., see 2.3), and the span of each range.