So, from the looks of it (I assume we are seeing RV870 and not RV840 there), RV870 would be a tweaked for DX11 RV790 @40nm, without a major increase in shader count (AT suggests something more than 1B trans).

If AMD goes again for best value/cost for the performance segment (employing the same binning scalability that the RV770 strategy used), they will have the same competitive products in the 200-300$ range, with better margins per chip (but less yields are implied, at least atm due to TSMC difficulties with the process).

Can that be enough to counter the Nvidia loch ness monster?...also, will it even matter for most users, since nVidia will target the 500$+ range?