🚀 go-pugleaf

>>: > I somewhere heard that ALL pentium processors which are made cost intel
>>: > less then $100 to manufactorer (without R&D-costs).....
>>
>>: As far as I know, they make them at below $50 per die. Add about $15 for the
>>: package -- and that's it: You can start selling this beast for 100s of $...
>>
>>: I knew that there must have been a reason, why CPUs exist...  ;-)
>>
>>What about the development costs. It costs huge amounts of money to
>>develop somthing like a Pentium. This cost has to be recovered by
>>then selling tham at well above the manufacturing cost.
>
>
>Let's do some arithmetic.
>
>Say the plant cost $1 billion (average for the .35 micron stuff) to
>build and has 5 good years before it is considered obsolete (and is
>worth only 1/2). The cost of money is say 10% per year to make it
>easy.
>
>This mean each year, the plant costs $200 million just sitting there,
>or 4 million each week. Assuming the plant can do 2,000 wafers each
>week, we are looking at $2,000 per wafer *before* the wafer arrives.
>
>Then you add in the raw material, the chemicals, the people, the
>packaging, etc. (And as Richard pointed out, the R&D), the support
>cost, the data books, the advertizing. And you can see that Intel
>has to work hard on this gold mine; the gold does not drop from
>heaven.

	Equipment and fab operations are the major costs. While photoresist,
30% peroxide, and wafers themselves might seem like a major hit (peroxide is
about $250 a gallon) those costs are 'diluted' by bulk processing like wet
cleans and so forth. Resist is expensive, but there's maybe 2 tablespoons per
8" wafer dispensed for each lithography run. Not much. Wafer starts per week
are meaningless unless you consider wafer diameter.

	On another note, the reason Intel is 'overclock protecting' is due to
certain resellers re-labeling chips as higher speeds. Keep in mind that the
chips are made first, then sorted according to highest safe clock speed. As the
process matures, the cost per chip decreases (you get more that work for a
given wafer) and the technology improves (smaller transistors=higher speed,
etc). Sooo, a third-generation 486/33 is probably able to run a lot faster, but
if there's a demand for a 486/33 it'll be sold as that, regardless of maximum
safe clock speed. This is where the 'protection' comes into play, since you pay
(Intel, for instance) for 33's you should only be able to sell them as 33's.
Otherwise they'd be losing money since a DX/2-66 should cost more than a DX-33.