Hans Meijer
Zoetermeer, The Netherlands
Moore's Law has been interpreted to mean so many things over the years that Gordon himself has admitted he's unsure of the exact meaning. His original 1965 article focused on the doubling of transistor counts every two years. If you focus on microprocessor performance, how
ever, the improvement is at the faster rate of doubling every 18 months because the increase in transistor count is abetted by an increase in clock speed, both of which contribute to overall performance. As this increase is driven by Moore's Law, it is treated as a corollary to Moore's Law, or even as Moore's Law itself. In fact, we combined the two variations in the article; we should have specified a two-year doubling when speaking of transistor densities. -- Linley Gwennap
The table "4004 vs. Pentium Pro" in "The Birth of a Chip" is confusing. If there are 2300 transistors on a 4004 with a 12mm
2
surface area (191 transistors per mm
2
), and 5.5 million transistors on a Pentium Pro with 196mm
2
(28,061 transistors per mm
2
), then the transistor density has increased 146-fold. At the same time, the feature size has decreased from 10 to .35 microns, or about 28.6 times, so the feature area (10
2
vs. .35
2
) should be over 800 times smaller. W
hy then has the transistor density increased only 146 times?
Daniel Cincunegui
dcincu@cadxpress.com.ar
Early chips such as the 4004 had relatively simple designs that were laid out entirely by hand, resulting in little wasted space between the transistors. Modern microprocessors require more space devoted to routing -- the interconnects between transistors -- and are typically laid out using automated tools, which are not as efficient as hand routing. These factors have prevented transistor density from simply tracking the square of the transistor size. Note that the Pentium Pro cache chip, which uses a very simple design laid out by hand, crams 31 million transistors into a 242 mm
2
die, giving it a transistor density 670 times that of the 4004, closer to the 800x improvement you calculated. -- Linley Gwennap