Originally Posted by Majin SSJ Eric
I'm not necessarily buying it either but you gotta admit it's a really cool idea and would drop like an atom bomb on the gpu scene. It would be a clever innovation and something to get people talking about AMD. Plus, you still could use the individual chips in 470-esque sku's while catching everybody by surprise with a legitimate 1080 competitor.
It's total fan fiction but it's a fun idea!
Your idea of an atom bomb for the GPU market is a bit much.
Although a different method, the performance and similar idea have been used in the past. That is to use 2 midrange chips on a single card to reach high end performance. The results and sales have generally been the same and that is generally an unsuccessful product as far as sales go. This would not be any different because CF has awesome scaling when it works. The problem is the software support and that would not change with two GPU on a single interposer. All it would do is create nightmares for RMA's, more difficulties in manufacturing and increase the failure rate of the product. One chip failing, would mean scrapping the whole card. Looking at past history is why this product wouldn't do as well you might think.
The prime example of this is Voodoo technologies last products or voodoo 5 and any videocard from asus which tries to make a dual GPU card with midrange chips. All these chips were priced around the price of a single GPU flagship or less but offered better performance overall. The exception to the rule that sold well was the gtx 690, but this was a flagship x2, which is people in the dual card market want.
Personally, I might consider buying one at 500 dollars, a dual Polaris 10 I mean. Certainly a more likely purchase than 700 dollar founders edition 1080, but I don't the market would accept it very well, particularly if the leaked 3dmark scores are true. Most people who want to buy a dual flagship class GPU card, not a dual midrange to get to flagship.
The problem is most people in the market for dual GPU cards are looking for better than single gpu flagship performance. And not just by a little bit, but alot. Otherwise products using 2, gx104 chips would have been more succesful aside from the gtx 690.
If AMD was capable of putting two 120watt product on an interposer, what would be an atom bomb is if Zen performs like haswell or better and AMD puts a polaris 10 chip on it and sells it for 300 dollars. That would be revolutionary. Better than console APU's is something people want and I am surprised AMD has not done yet.
Not 2 small dies performing like a larger single one or even out performing it. That has been done to death and isn't any more novel than the original Fury Nano design, which atleast had a unique niche.