Some Effects of LSI on Minicomputers
Total Page:16
File Type:pdf, Size:1020Kb
IMPACT OF LSi TECHN'ILOGY 3N SLIDE RULES. CALCULATORS, AND MINICOMPUTERS SOME EFFECTS OF LSI ON MIRICOMPUTERS C. G. Bell L. Gale C. Kaman Carnegie-Mellon Univ. Digital Equip. Co. Digital Equip. Co. The minicomputer is entering the HOLD TECHNOLOGY CO?dSTANT era of the consumer good; it is forced by a combination of market We believe that the main effect of pressures (e.g. demands, and drastically reduced costs will be eventual desk calculator competi- the creation of a very large market. tion) and available low cost Already the most dramatic realiza- technology. If system costs are tions are the desk calculators and held low in the order of $10,000 the programmable calculators. to $20,000, then a number of functions can be incorporated to The view of minicomputers is chanq- aid the user. (This means that ing to handle the control within the mini is an important contri- various terminals. Observe the use butor to the evolution of computer of minis in terminals--the so-called systems.) Thus, lower cost and "smart terminal": for remote batch higher performance technology pro- processing terminals with included vide three possible alternatives: card reader, punch and line printer (e.g. the UCC terminals and DEC DC71): 1. Hold technology constant and for one-user graphics terminals (e.q. provide the minimal cost system the DEC GT40 and ~T08);and finally, for that technology (in essence, we might expect that lower cost soft the factor creating the mini- and hard copy terminals would be computer). controlled in a similar fashion. Indeed, eventually yesterday's minis 2. Hold cost constant (at what priced for the consumer market will users might pay) and incorpor- be at today's calculator prices. ate as many functions as Yesterday's large, general purpose possible for the constant cost. computer is today's minicomputer. Dennis and Smith (wl), of IBM 3. Build structures which would research, postulate a 5000 circuit not have been possible with and 20,000 byte computer integrated older technology (very large circuit cost in the late 1970's of array and pipelined computers about $75. such as ILLIAC IV and CDC STAR, see Bell, Chen, and Rege, Now, almost every integrated circuit 1971). manufacturer has an MOS "micro- processor-on-a-chip" (e.g. Intel The following discussion explores 80081, and by combining several dozen the first two topics. chips together with memory, a 5 -U 10 TECHNICAL SESSIONS microsecond cycle time minicomputer programs larger than the physical can he fabricated. memory. This capability will no doubt be standard in five years. For a one user system with a key- board-printer (or scope) terminal, In ,the future, other capabilities the system price is beginning to be like those above, which have been dominated by the mechanical parts derived from larger, general purpose of the terminal. For a small volume, computers (see Bell and Newell, 1971) the hardware and software costs are will undoubtedly be available, for about equal. For the user, the example, the use of caches to increase applications' costs predominate. In performance. fact, in an article (Coury, 1970) a "0" cost minicomputer was predicted, Although the above concepts have been and still the system cost was over "adopted" from the large computer, $10,000. the minicomputer has also contributed to computer development, Perhaps the HOLD COST CONSTANT greatest influence has been to belie the myth that computers are hard to From a computer structures and user interface. The research necessary viewpoint, the addition of more for multiprocessing is based on mini- functions in a minicomputer has more computer processors (e.9. Wulf and promise. The total cost to buy; Bell, 1972). Similarly, user micro- maintain, program, and use the mini programming is available on several is important--not the purchase cost minicomputers. This capability is alone. Admittedly, the manufacturer douhle edged: while giving higher has difficulty communicating this to performance, it does so at losg of buyers who tend to look mostly at generality and future compatibility. purchasing price. OBSERVED EFFECTS SO FAR For example, a larger primary memory provides probably the largest incre- Already the minicomputer, because of mental improvement in cost-effective- its simple structure and relatively ness because it permits a larger short half life, has tracked tech- range of user and systems programs nology. Thus, any change in circuit to be run. performance or availability is in- corporated rapidly. The advent of There are other improvements already LSI has effected the cost and market appearing in the more advanced mini- size as described above. computers. The processing of float- ing data in hardware provides the The direct effect of low-cost, high- single most direct improvement in performance LSI memories has per- performance. With floating point mitted a large processor state (i.e. capability, there is less distinction general registers); microprogramming between the larger general purpose (1/2K word read-only or read-write computer (e.g. mid-range 360 and 370 memories) ; various memory mapping models) and the minicomputer (e.g. (constructed of smail arrays); DEC PDP-8 with floating point pro- eventual replacement. o:~core by MOS; cessor, Interdata computers, and and ?he cache memory. Within the PDP-11/45) . Memory mapping capa- processor tl19re has been a less bi Zity (e. g. procrram paging and/or dramatic effect; an MOS processor-on- segmentation) also provide for multi- a-chip is really only cost-effective programming .and the execution of in smaller systems where absolute IOW MEREM 72-151 cost is required--or a certain minim.^^. perfom~3nc.r.-2s r;c~eptaSle. FOP example, the availability of the MOS PDP-8, runriny 3t 1/5-1/8 the current PDP-8 speed wodd not satisfy all the PDP-8 users. Instead, it would probably establish varkets which currently are not available because of cost. SUMMARY The above discussion has explored the technological (LSI) and market pressures (lower cost and more capability at a constant cost) which have influenced the minicomputer. As technology improves, it is in- corporated within the minicomputer so rapidly that the mini. now plays an important role in the evolution of the computer specie. REFERENCES Bell, C.G., R. Chen, and S. Rege, "The Near-Term Effect of Technology on Computer Structures". Bell, C.G., and A. Newell, Computer Structures: Readinqs and Exampls, McGraw-Hill Book Company, 1971. Coury, F., "A Systems Approach to Minicomputer 1/0" SJCC, 1970, p677- 682. Dennis, S.F. and M.G. Smith, "LSI: Implications for Future Design and Architecture", IBM Research Report, RC 3598 (#16278) November 3, 1971, IBM Research, Yorktm Heights, N.Y., 26 pages. House, D.L. and R.A. Henzel, "The Effect of Low Cost Logic on Mini- computer Organization" Computer Design, January 1971, pp 97-101. Wulf, W. and G. Bell, "C.rmr(p: A Multiple Minicomp~!ter (pDP-ll) Processor" FXC, 1972. TECHNICAL SESSIONS .