% fortune -ae paul murphy

Moore's Law, RISC, and mis-understanding

Here's the original, 1965, statement of what became known as "Moore's law"

The complexity for minimum component costs has increased at a rate of roughly a factor of two per year. Certainly over the short term this rate can be expected to continue, if not to increase. Over the longer term, the rate of increase is a bit more uncertain, al- though there is no reason to believe it will not remain nearly constant for at least 10 years. That means by 1975, the number of components per integrated circuit for minimum cost will be 65,000.

In reality, however, Intel's own 8086 CPU had only 29,000 transistors when released in 1978 and today's Itanium 2, counting all 9MB of on board cache, only has 592 million - less than one ten millionth of the number predicted by an annual doubling over the 40 years since Moore's comment.

If, however, Intel's megahertz race had started in 1978 and followed Moore's law as widely misunderstood to say that nominal clock speed doubles every eighteen months, today's 8086 machines should be running at about 98 Ghz -an order of magnitude greater than Intel's current best, but in the same ballpark as Sun Niagara's if we admit parallelism.

In reality, however, Moore's observation was as much about cost as it was about component densities since he was really describing a relationship between the two. Step outside the cost limitation and considerably more has been achieved.

By 1969, for example, Ray Holt and his colleagues at AiResearch had developed and tested the CADC microprocessor board for use in the F-14 flight control system. The CADC single board computer, whose weight and power gains over pre-existing digital machines were on the same scale as those being achieved today by both IBM's Cell and Sun's CMT initiatives, incorporated a 20 bit, 375Khz, CPU whose performance and reliability Intel wouldn't match for at least another ten years.

The key thing Holt's CADC, IBM's Cell processor, and Sun's CMT initiative have in common is their role as the first hardware implementations of advanced software and design ideas -i.e. the revolutions these things kick off aren't as much about transistor densities and hardware as they are about ideas and software. Thus the CADC was far more about advancing the art of software mediated control systems than it was about hardware, IBM's Cell processor is only the "prefered embodiment" of a set of ideas about managing process communication in a computing grid, and Sun's CMT is fundamentally about moving the cost curve for SMP processor management downward by expressing key ideas directly in the hardware - and thus ultimately about bringing the evolutionary power of the marketplace behind Moore's observation to bear on entire systems, not just processors.

Paul Murphy wrote and published The Unix Guide to Defenestration. Murphy is a 25-year veteran of the I.T. consulting industry, specializing in Unix and Unix-related management issues.