Section 1
The IBM 701-7094 II sequence, a family by evolution
The IBM 701, 704, 709, 7090, 7040, 7044, 7094 I, and 7094 II sequence relationship is shown in Fig. 1. The group is not a compatible series. The IBM 701 [Astrahan and Rochester, 1952; Buchholz, 1953] is a forerunner of the series; all except the 701 are painfully compatible. The sequence is included because the 7090 is a reference or benchmark of scientific-computer power. All machines use 36-bit words. The 701 stores two instructions/word in the same manner as the IAS computer (Chap. 4), whereas all others in the sequence store only one instruction/word. The 701, 704, and 709 are first-generation, vacuum-tube technology; the rest are second-generation.
The IBM 7094 II description given in Chap. 41 is based directly on information in the Programming Reference Manual, but the Appendices of that chapter give the ISP of the Pc, a Pio, and a K as inferred by the authors of this book. The description of the Pc gives the instructions in the 704 and 7044
Fig. 1. Relationships among IBM 701, 704, 709, 7094 series.
Fig. 2. IBM 701 PMS diagram.
to show an evolution. However, the major evolutionary change does not appear in Pc's ISP but in the PMS structure.
The 704 structure, like that of the 701 (Fig. 2), provides only for peripheral transfers to primary memory via Pc under programmed control with no interrupt system. As such, only one T or Ms could operate easily at a time. The 709 introduced the Pio(' Data Channels) to improve the ability to transfer data between Mp and Ms without requiring Pc intervention. Concurrent operation of several I/O devices is carried out by multiple Pios along the lines of the 7094 II PMS structure (Fig. 1, Chap. 41, page 518). However, the utilization of the data channels tends to be rather low, particularly when the data channel is controlling very slow devices (e.g., card equipment and line printers). When operating a high-speed tape unit at 90,000 x 6 bits/sec the utilization of the data channel is still only approximately 3 percent. A program interrupt method of data transfers would have been sufficient.
The incompatibility among the machines, especially the 7090-7040-7094, is disheartening, both from the point of view of a user and an engineer. The incremental hardware needed
515