The microprocessor rationale

Bruce Threewitt
1975 Proceedings of the May 19-22, 1975, national computer conference and exposition on - AFIPS '75  
THE MICROPROCESSOR RATIONALE Classically, since the beginning of the computer era, logic system designs have been implemented with one of two entirely separate approaches. Economic factors determined whether hardwired logic or a computer would be used to perform the required operations. In many cases, the computer approach was far too costly. Until recently, hardwired logic systems were constructed using large numbers of small-scale or medium-scale integrated circuits (SSI and MSI,
more » ... d MSI, respectively). Figure 1 includes examples of functions typically found in these two levels of integration. In general, the hardwired logic system could be modelled as a "black box" that takes inputs (variables/data) and performs some defined logic function on them, producing outputs (results) that are functions of the inputs and of time (see Figure 2 ). These outputs drive displays or other output devices that apply the results to doing work. Meanwhile, the computer industry builds black boxes that accept inputs and perform some defined logic function to produce outputs that are functions of the inputs and of time ( Figure 3a ). The difference in these approaches lies in the contents of the respective black boxes. Computer systems have a certain minimum configuration as shown in Figure 3b . Until now, this configuration consisted of components and software that were too expensive to use in the simpler high volume logic system applications. As mini-computers tend downward in size and cost, more logic applications can be served; but, the cost/performance trade-offs of mini's still preclude their use in general hardwired logic replacement market. As the need for less expensive, more versatile logic system DISCRETE SSI MSI LSI ~ ~ I COUNTERS I CPU g:::Qo-o i ~ 1 I REGISTERS I I o---t>o---o Figure I-Trends in integration 3 design techniques continued to apply pressure to the digital electronic industry, a third system design alternative was evolving for the following reasons. First, logic designers were frequently going toward large-scale integrated circuits (LSI) to implement their particular system designs. Typically, a logic designer would take his logic and timing diagrams to an integrated circuit (IC) supplier and ask him to shrink that design onto one or a few MOS/LSI circuits (see Figure 4 ). He did this because he believed that the MOS/LSI technology offered a high-density, low power approach to digital circuits. Unfortunately, this assumption is not totally valid. It is true that the MOS/LSI technology yields small transistors. However, the MOS interconnection approach is significantly less dense than cells or devices. Random logic is characteristically dominated by complex interconnect with relatively few active devices (transistors). Thus, random logic designs cannot utilize the MOS technology efficiently. MOS/LSI does make very dense memories (patterned logic) where interconnects are minimized. Secondly, the cost of producing custom MOS/LSI circuits is prohibitive in the long run both for the user and the IC manufacturer ( Figure 5 ). This curve assumes that for random logic, increased complexity (integration level) results in increased specialization. When a logic system, which is specific to a given special purpose application, is integrated, the resulting Ie's are also specific. Thus, the volume per product type decreases with increased integration. The few exceptions to this rule are single-application random logic devices such as calculator and watch circuits that generate sufficient volume to justify their existence. Also, the IC supplier must supply many different types of custom circuits in order to adequately serve the custom logic market, since any two given applications would probably not utilize the same custom circuits. This proliferation of circuit types greatly increases overhead costs for support engineering for testing and circuit design. Therefore, custom MOS/LSI, in general, is not economically feasible for the Ie manufacturer or the systems manufacturer. Thirdly, the appearance of the N-Channel silicon..,gate MOS technology in recent times provides a vehicle for OUTPUTS = f (INPUTS) (RESULTS) Figure 2-Logic system implementation I From the collection of the Computer History Museum (www.computerhistory.org)
doi:10.1145/1499949.1499953 dblp:conf/afips/Threewitt75 fatcat:y4wxzlvnrvhr5ifedffm46cjae