Originally posted by: invidia
So the smaller chip agrument is just for cheaper costs?
That is mostly it.
1) Smaller chips = more products per silicon wafer. That means with the same materials cost, the same labor, the same machine time, you get more chips. For example, moving from .35 micron sizes to .25 micron sizes means you get double the number of chips per wafer (I know it isn't exactly double but lets not complicate things). Going from .25 micron to .18 microns gives you another double. Going from .18 to .13 doubles it again. In the end you get 8 times as many of the same chip. Thus, costs are 1/8th the cost of a large chip. When we are talking about a $50+ CPU, 8 times the price is a big deal.
2) Errors. There are flaws in the chip manufacturing process. The flaws will be fairly randomly distributed (I know there is a bias towards the outer edges, but lets not complicate things). If you have a small chip, the chance of a flaw being on the chip is small. If you have a very large chip, the chance of there being of a flaw on the chip is astronomical. Basically, if the chip is too large, you'll be virtually guaranteed an error somewhere. An error which may very likely make the chip unusable. If the chip is too large, you throw away a ton of chips because they don't work. Take that 8x the cost and make it much higher.
3) Geometry. Chips are roughly rectangular. Silicon is a circle. Try taking out a piece of paper and draw it. Draw a circle. Put several large rectangles inside it. Put as many large rectangles as you can. See all the waste along the circle perimeter? That is wasted materials, wasted labor, and wasted machine time. Try it again with small rectangles. See how the waste is drastically smaller? Take that 8x the cost for a chip and make it much higher due to waste.
The rest is just user friendliness. You don't want that IPod nano to be the size of a football do you?