174 BEGINNING OF THE MINICOMPUTER
precision and color cathode ray tubes. The PDP-l, 4, and 7 were relatively compatible in terms of I/O interconnection and evolved to have about the same set of options. PDP-9 changed to an I/O bus structure, requiring new option interfaces. Although PDP-15 used that same I/O bus structure and signals, the voltages were different; again, new option interfaces were required.
Displays have been major options throughout the series. Moving head disks were first available on the PDP-15. Although a number of card handling options were available, few were sold, reflecting the real-time, laboratory, and multiprogrammed (timesharing) use.
This chapter concludes by relating the 18-bit series evolution to the model of minicomputer evolution presented in Chapter 1. Three design styles are distinguished in the model, as can be seen in Figure 42. Chapter 7 shows the 12-bit family (PDP-8) evolving mostly along the constant performance/decreasing price curve. The 16-bit PDP-l 1 family, presented in the chapters of Part IV, evolved based on all three design styles.
Figure 42. Three design styles.
For a family to evolve in more than one design style, design resources must be available for parallel development efforts. While the PDP- 11 family had the multiplicity of designers and architects to do this, the 18-bit series did not. Each new implementation was designed by a member of a previous implementation team. For such a single-thread approach to be successful, it appears that one of the three design styles of the evolution model must be chosen and consistently followed. With the exception of PDP-4, the 18-bit series has followed the-middle style: constant price/increased performance.
It appears that a clear identity is needed to guide design decisions. Consider the physical packaging of the last of the 18-bit machines, the PDP-15. Although a comparable speed/ performance PDP-l 1 required more integrated circuits to implement (the PDP-l 1 has more modes of addressing, more instructions, and more data-types), the PDP-15 implementation cost more. The PDP-15 remained packaged in a large cabinet, used smaller modules, and the component density per module was lower than that of the PDP- 11. Had the evolution been guided by a consistently lower cost goal, metal box packaging rather than cabinet packaging would have been used. As it was, the PDP-15 had to compete against the PDP-l 1 with the handicap of an extra level-of-integration in its physical packaging.
Several people helped gather the data for this chapter and critiqued its design: Dick Best, Earl Cain, Wes Clark, Dick Devlin, Craig Mudge, Carl Noelcke (reliability calculations), Ed Rawson, Jack Shields, Dan Siewiorek, Don White, and Don Zereski. Mary Jane Forbes and Louise Principe deserve thanks for typing the numerous drafts.