i am saying that it is very large for a 28nm midrange chip
And doesn't it support Quad-SLI?
i am saying that if they need 300W for a *midrange* 28nm chip, then they are in trouble
(which should indicate to you that i don't think this is the GK104)
Wait a minute, the picture shows 2x 6-pin connectors, meaning 225W max.
However, it does seem kinda "fake" because of the PCB power connectors - it looks like the stacked-up power connectors were just slapped onto top of the board PCB, disregarding the actual holes). Maybe NV is just providing the flexibility to AIBs with the reference PCB so that they could use different cooling solutions without having to use the stacked connectors (and a 6 + 8-pin possibility for OC'ed editions that would need a bit more than 225W).
Calling it "midrange" would be like calling 7900GTX "midrange" at launch. The 7900GTX was a pretty small and cool-running chip - I know because I've done custom cooling on it with a CPU heatsink - it was tiny enough to perfectly fit the copper fins of that Zalman flower heatsink.
Why do you act surprised that the monster Kepler will be over 500mm^2? GTX 280 was 576mm^2 (and a 55nm 9800GTX+ was released shortly afterwards, while it took Nvidia forever to shrink GT200 down to 55nm).
EDIT: The size of the picture looks like 320 to 340mm^2. One guy said that it's 18x19mm, which is 342mm^2, but others are saying that it looks a bit smaller than that. That's actually smaller than GF114 (360mm^2).