MadSci Network: Computer Science |
Sarah,
"Moore's Law" is not really a "law" in the sense that Ohms Law (i.e., that voltage = currant x resistance) or the law of gravitation are. It's a prediction about how quickly the the number of transistors in the central processing unit (CPU) of computers (microprocessor "chips" in computers) grow over time. As your researches have turned up, it is based on Gordon Moore's observations on the subject, about thirty years ago. Exactly what those observations were, I don't know but the outlines of what he did are fairly clear. He observed that over the years the number of transistors in a "typical" CPU was going up. Then, being familiar with exponential growth, he wondered if this growth followed an exponential pattern: If the number of transistors in a typical CPU on some date is To and the number n years later is Tn then
Tn = To x 2(n/d)
for some d, the doubling time in years. When he originally looked, choosing 2 for d let the equation fit his data pretty well. Later, when there was more data (and after the pace had picked up), it became clear that 2 was too big, so he changed it to 1.5 (i.e., 18 months).
Like all technology, computer technology is obviously a human endeavor that takes place in the world around us. As such, part of what Moore's Law is trying to predict is determined by things like how hard people work on improving the technology. Since that could change (working on computers could fall out of fashion the way working on space exploration did after Apollo), the speed of CPU development as measured by Moore's law could change, thus invalidating it in a way that Ohms law never will be. Even so, Moore's law has been a reasonably accurate predictor for some thirty years. A pretty remarkable result considering how little data he had to begin with -- which is why people still talk about it.
You also asked for some raw numeric data. While these are obviously not the data Moore originally used (since the chips listed weren't invented when he wrote his "law") they are typical of what people use to demonstrate the "constant" doubling time.
Year | Intel Processor type | Number transistors (approx.) |
1981 | 8088 | 29,000 |
1984 | 80286 | 134,000 |
1986 | 80386 | 275,000 |
1990 | i486 | 1.2 million |
1993 | Pentium | 3.2 million |
1997 | Pentium II | 7.5 million |
Hope this helps.
Try the links in the MadSci Library for more information on Computer Science.