Vinod Grover's Observations re. Proebsting's Law

Vinod Grover's Observations re. Proebsting's Law
(May 31, 2001)

I've been looking at publicly available data to see historical support for Moore's law and Proebsting's law. I found the published specint95 records from Aug95 to jan 2000 for various vendors and compilers to see if I can mine some trends in there. I haven't finished analysing it yet but I discovered interesting datapoints that cast some doubts on, both, Moores law and yours.
Here are two data points concerning Proebsting's law.

HP published integer results for model J210 (100 Mhz PA RISC) in Aug 95 and Aug 96 using two different compiler versions (10.13 and 10.32 respectively). The benchmarks were identical in both cases. The specint results were 4.37 vs 6.0 which is roughly 37% improvement in 1 year.
(See http://www.spec.org/osg/cpu95/results/res9509/p030.html and http://www.spec.org/osg/cpu95/results/res96q4/cpu95-960903-01253.html . )
Later HP published results on identical machine Model b1000 (300 Mhz PA-RISC) in feb 99 and Jul 99 using two different compiler versions 10.21.20 and 11.01.00 respectively. The published spec results were 24.1 vs 22.4 which is roughly 7.5% in 5 months.
(See http://www.spec.org/osg/cpu95/results/res99q2/cpu95-19990524-03664.html and http://www.spec.org/osg/cpu95/results/res99q3/cpu95-19990816-03767.html . )

Vinod Grover's Observations re. Proebsting's Law (May 31, 2001)

Vinod Grover's Observations re. Proebsting's Law
(May 31, 2001)