2. Moore's Law is packed with economic issues. For example, Stokes describes the problem of maximizing the efficiency of chip production, and the basic trade-off involved.
in between the extremes of a highly integrated but wasteful single-component design that puts all 10,000 transistors on one large/complex chip and a less integrated but more expensive multi-component design that distributes the transistors among four small/simple chips tied together on a PCB, is there a number of chips and a level of complexity per chip that costs the least to make? Ultimately, we're trying to decide on the most cost-effective and least wasteful way to partition the 10,000 transistors in our processor design among one or more chips. What level of integration gives us the most bang for our buck?
What Moore noticed was that the rate at which engineers were discovering ways to shrink components was causing the minimum cost design to include roughly twice as many transistors per chip every one or two years.
Another issue with which Stokes deals is the fact that the demand for transistors-per-chip may be slowing. However, he argues that this means that the fundamental shrinkage of components is being used to fuel smaller computers or ordinary objects with computing capability (so-called pervasive computing). Thus, Moore's Law continues to be an economic force, even though it may not lead to building chips with a bazillion transistors.
For Discussion. Is the shrinkage of computer components best analyzed within the framework of exogenous technical change or endogenous growth?