It only makes sense that all NV3x cards will be based on the GeForce FX (meaning all programming-side capabilities of the GeForce FX, if not more).
I also "expect" any NV3x card to be based on CineFX architecture. However, if we continue with that assumption, that means that nVidia will still be doing something they never have done before. IIRC, nVidia has never used three separate chips from the same 'generation' to target mainstream (NV34 - $100), Performance (NV31 - $200) and enthusiast (NV30 - $300+) markets at the same time.
In the past, nVidia has always used two chips, with multiple speed grades / memory configs on each, to cover the three markets.
Using three chips would be a new direction for nVidia, so I would not "rule out" the possibility that nVidia is doing something "new" with their chip code names....possibly being NV31 and/or NV34 being DX8 based, with some features borrowed from NV30. Having three separate chips with similar feature capability in each market sounds like an expensive and risky way to go, especially if they are available at nearly the same time.
The NV31, therefore, will likely be a cut-down version of the GeForce FX...
Or, it could be a "built-up" version of the NV34.
The $200 price point is a risky proposition for all IHVs. There are three ways to approach it, and choosing the wrong one can get you into trouble:
1) Use the same chip as the enthusiast market, but cut down the board level product some way....Lower clocks, narrower bus, less expensive memory, or ATI's new approach of 'disabling pipelines'.
2) Use the same chip as the value market, but improve the board level product some way....Faster clocks, wider bus, more expensive memory.
3) Use a completely different chip.
As with most things, there are pros and cons with each approach:
Approach 1 will probably get the best product, but with the lowest profit margin. (Better for consumer, worse for IHV). Think Ti 4200.
Approach 2 is just the opposite...worst product, best profit margin. (Worst for consumer, best for IHV). Think GeForce4 MX 460.
IHVs would prefer approach 2, but competition might force the path of approach 1.
Approach 3 should lead to the most balanced product...but there's a much larger "up front" R&D effort involved in fabbing a separate chip, testing, driver support, etc. (Vs. minimal R&D effort to upgrade or downgrade a board level product.) You also run the risk of having "too many" products in too narrow a price range, particularly if the enthusiast product must be forced down in price due to competition.
I'm just not sure that I see room for three separate NV3x chips, all with similar levels of feature (API) support, and selling within $200-$300 of one another. Not sure nVidia's OEMs would appreciate that.
Having said all that, it looks to me like ATI will be essentially using approach 3 come the spring. I'm guessing RV-350 chip at value, R-300 chip at performance, R-350 at enthusiast. The only difference between this, and an "all DX9 NV3x" situation, is that the R-300 will not be a "new" chip when RV-350 and R-350 launch. That gives ATI quite a few more viable options with how they decide to fix their line-up in response to competition.
