Mianca- You are set on your POV of the rumor mill, just don't be surprised when your speculation turns out to be completely wrong.
Sorry if my comments appeared a bit opinionated - it's just the way I'm used to make my points clear. That doesn't mean I don't respect other people's views on the "speculation game".
Barts is a good bit cheaper to make than a 5850, not taking into account the memory ICs. Similar/same PCB but different, probably cheaper, components, smaller die by ~15%, uses similar or less power.
Yeah, and that's exactly why I (personally) think it would make a great HD 5850 replacement. Better value for the customer, more profits for AMD, everyone wins. I'm well aware it won't work that way if Barts XT isn't somewhere near Cypress XT in performance, of course. So that's where my entire speculation hinges on.
Now I wanna play the speculation game:
Barts XT - 6770 - ~140-150w TDP, ~270-280mm2, competes with GTX470.
I was originally thinking the MSRP would be ~$249 but that seems a little low, $269-$279 sounds a bit better for something that performs between a 5850 and 5870.
Pretty well. Looking at my own share of speculation further above, the only point in which we actually seem differ is the naming/price tag. You say they'll sell Barts XT as HD 6770 @ $279, I'd rather suggest they push it into the high-end segment, label it HD 6850, and make some extra money (>=$300).
Time will tell who's right. In the meanwhile, there are arguments in favor of both views. I stated those in favor of mine rather clearly in my previous posts - but the main arguments actually are
(a) profit and
(b) marketing:
I already talked about
(a) further up in this post: Rather sell a card that is cheaper to make than your last HD5850 card for the same (or more) money in the high-end segment than keep that same card in the high-midrange segment and use it to replace the (vastly cheaper to make) Juniper card. The latter would basically be great for customers (HD5870 performance @ HD6770 price), but bad for AMD's profits. You decide what a highly indebted company probably has to care more about.
As for
(b): People who already own an HD 5770 simply will be quicker to pick up Barts XT if it's labelled as "HD6850" than if it's labelled as "HD6770". Moreover, there's the psychological aspect of evoking the impression that basically, your cards are all about high-end performance: Rather make it look as if you "add" performance on top of what already is on the market (i.e. shift your entire lineup one "half-step" towards the enthusiast, HD69xx level) than sell your stuff cheaply. If your new upper-high-midrange card can actually compete with your competitor's lower-high-end card (GTX 470) - why not label your upper high-midrange card as lower high-end (HD x850) card straight away?
Everything that's better than your competitor's best should actually be placed no lower than the HD69xx category - so that's where Cayman XT should have to be placed given current performance rumours; and as for the gap created below Cayman Pro, well, that's where Barts XT would just fit in.
On top of all that, there's
(c) the power consumption aspect Squilliam pointed out once more (if you compare his suggestions to mine, you'll see that I actually support his view - with Barts XT @ two 6pin power connectors just not fitting into the high-midrange anymore practically being my very first reaction to seeing the leaked picture of it).
When all is said and done, everything I suggested might very well turn out to be false. But I still think that,
provided that my basic performance assumptions for Barts aren't
totally off, the lineup I proposed strikes a very good balance between (a) performance gain over the last generation cards at the same naming level, (b) categorization with respect to power consumption and (c) price targets actually making the launch of those new cards financially worthwhile for AMD.
Finally, here's an updated version of the PCB comparison I made:
HD 5770
Barts XT
HD 5850
HD 5870
You be the judge on whether it makes sense to sell that complex Barts XT card @ a high-midrange price.
No, no, no. How many times are you going to write this thingie? 
