>But what I still don't understand is the constant release of faster chips incrementally. Of course it is how these chip companies make all their money, but it has to be one of the biggest scams ever.
>
>Do you actually believe that every couple months they invent these new techniques for making the chips faster? They are just milking us for all we are worth. I bet if they really had to, they could create a 1.5g chip right now. But why should they when we'll by chips for every 33mhz interval that they manufacture along the way.
>
>Am I missing something?

Yep - there are constant improvements in substrate and masking technologies, which allow more stuff to be packed into less space, produce less heat, and give a higher yield at any given resolution. I have a couple of friends who work on some of the hardware used to actually burn the chips, and they hardware is constantly being revised, tweaked and improved.

The newer chipset from Intel are using the Coppermine technology, which uses both new substrate and packs the components more tightly on the chip. besides new chemistry, the lasers used to etch the chips all had to change their operating frequencies, the tolerance of the electromechanical parts, even the air tables that hold the wafers in place when the circuits are being etched had to change. IBM introduced the use of copper rather than aluminum in the substrate a while ago, but it's taken time to get the chip yields up to a level where its economically feasible to sell them.

I'm sure that several of the big semiconductor vendors can produce higher speed chips, but if only one in a thousand will operate correctly at the targeted speed, and it requires liquid nitrogen to keep the chip from slagging, it's no good other than for bragging rights and curiousity. And as speeds go up and memory sizes increase, the tiny delays in getting electrical signals to move a fraction of an inch become more and more significant, so that even if the chip can run faster internally, the signals coming from RAM and the chipset can't get to the CPU that quickly. While we've seen internal speeds increase rapidly, the external bus speeds have remained relatively slow - it wasn't that long ago that the external bus of eventhe fast processors ran at 66MHz, and the increment in the new chips, going from a 100MHz to a 133MHz front-side bus (the speed that the chip interfaces to the outside world) isn't anywhere near internal speeds, and memory tehnology had to change in order to make the shift to the 133MHz clock - PCI 100 SDRAM just couldn't get the data there fast enough, and the cost of memory has gone up recently, as less SDRAZM is being produced and more RDRAM is needed...

I'd guess that we're not really CPU-bound now as much as memory and I/O bandwidth bound, but without doing a detailed study, i can't be certain. there are a number of web sites you may want to check out, sysdoc.pair.com and www.x86.org are good starting points for the hardware and CPU instruction sets respectively.