GK110's price is not only demand driven. It is largely supply driven. It is practically at least twice harder to produce since it not only takes about double the space on a wafer, it also fails more often. NVIDIA might have the problem oil producers have now. Yes, in limited markets they may get a boost but after a point they loose customers in whole if they reach alternatives with a lower price.
So, I also tend to think of thoughts "lower price due to competition" but that might only happen if TSMC drops prices (the theory is that they gradually do, but that's also depended on their decision and other factors) which means AMD may be unable to drop prices if they did the same or NVIDIA might be unable to drop prices with any discussed solution, unless they want to go Kamikaze (a suicide mission) for a chance to get a market share, which is practically what Console makers are doing with their ludicrously cheap hardware but at least they have the backing of an entire collection of software bound to them, so they can do it.
Originally Posted by metal_gunjee
I just now found this thread and I haven't read every post, so forgive me if this has already been said and/or debunked...
I think it would be far more likely for a dual-GK110 card to have somewhere between 2048-2304 shaders for each GPU rather than it being a dual Titan.
GK110 is already a big die in industry's standards. It already is considered novel, high price and most probably failing on manufacturing more than other chips (it is most likely the reason 780s exist in the first place).
Can they do it? Yeah, but enlarging it further gives the idea it will probably give an exponentially higher fail rate. If it's now around 20% it may skyrocket to ridiculous rates.
And that would might mean an exponentially higher price.
Big progress will come with 20nm, not now.
Now is for optimizing current solutions.