mark@mips.COM (Mark G. Johnson) (01/10/90)
In article <Jan.8.22.57.41.1990.234@klaatu.rutgers.edu> josh@klaatu.rutgers.edu (J Storrs Hall) writes: >I'd like people's reactions to these predictions on microprocessor >technology by the turn of the century. They are intended to >indicate something on the order of "state of the art but commercially >available", use the "1990" column to guess what level I'm getting at. Have a look at the October, 1989 article in _IEEE_Spectrum_ magazine, entitled "Microprocessors circa 2000". It predicts that in the year 2000, Intel will claim 700 MIPS performance for its microprocessor (fig. 1). I predict microprocessors will deliver 1056 SPECmarks in A.D. 2000, where SPECmarks is the ratio of the uP's runtime to that of a DEC VAX 11-780, on the large SPEC benchmark suite. This comes from applying the formula {2X every 2 years} to the starting point (1988, 16.5 SPECmarks). Initially this it is *conservative* in that it predicts 33 SPECmarks for 1990 whereas, I'll bet you 5 rounds of beer, SPEC will publish measured results in 1990 of microprocessors with >>40 SPECmarks performance. Note: SPECmarks are generally lower than the vendor rated "MIPS". So my prediction of 1056 SPECmarks means in essence, well over 1000 MIPS. Micros are approximately 12 years behind mister S. Cray in clock rate. (Cray-1 had 80MHz in 1977, micros had it in 1989 (BIT ECL SPARC).) However, the clock rate of micros is improving more rapidly than that of Mr. Cray's machines. Therefore, I'll say that Cray will ship the 500 MHz Cray-3 in ~1992, and a 500 MHz micro will ship in the nineties too. Obvious disclaimer: the above remarks are *my* opinions, not those of my employer or its liability insurance company. -- -- Mark Johnson MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086 (408) 991-0208 mark@mips.com {or ...!decwrl!mips!mark}