At 6800 release, it was $299 non-ultra 12-pipes, $399 GT, $499 Ultra. There were no $600 parts, and no 6600 mid-range parts.
Are you expecting 7600 to be 256-bit? Are you expecting it to have as many transistors as 6800gs? If you answer no to either of those questions, what are we arguing about other than semantics? If it turns out to be yes for both, then I'd be more inclined to agree with you --even tho 6800gs gpu has more than 20% more transistors than X1600 and even tho X1600 is on a smaller process (alas, B3D's table does not include die size for NV42 to make the comparison to RV530's 156mm2 --tho interestingly enuf, NV43, for the most recent "true" NV midrange part 6600GT is 150mm2, very comparable to RV530).
I've been saying for quite some time that it seems to me we are moving to a 4 part range from the old 3 part range because of the larger price range to cover, and the obvious add would be some kind of "super mid-range". But is there any evidence that NV will have 2 90nm gpus covering from $200-$400 at introduction, with another above that for the real high-end? I'm really curious about that for both players, if they officially go to 4 gpus, or try to cover in other fashions with last gens parts as NV is doing with 6800gs, and ATI seems set to do for awhile with X1800.
Again, I think 6800gs is a great value. And certainly it is not in the least "unfair" or anything else for NV to be doing --I'm just not convinced it is indicative of long-term trends yet. There are two very different conversations tho --one is the consumer angle, and wearing that hat all I care about is what the price tag and performance are. The other is more engineering/business, how many different asics and what are their costs of production. I suspect that the business guys would say in the long run both angles need to line up relatively closely, but in the shorter run there is room to make the needs of the 1st the master of the 2nd.