Yes this is exactly how I feel it's going to play out, if they have cracked the 28nm process and don't have to fuse off too many things on faulty GPU cores then they may show amazing performance. BUT and the real BUT is can they make these cores 100% working or even 90% working to make a high end card that shows a nice jump in performnce relating to a new process.
Lets see really, AMD may find a large percentage of their cores are faulty to the point they have to sell them as low end or mid range cards and this may help with prices in that range of cards for the consumer due to AMD having many bad cores on their wafers and instead of trashing them they will fuse off the damaged parts and sell them as cheaper cards.
Most of this just confirmed you have no clue what you're talking about, you don't make 50million of wafers on a whim then find you can only sell them as low end dies.
You don't make a 320-380mm2 die, and sell it as fast as a 80mm2 one. This categorically will not happen at all. nor are there any signs of process troubles.
Sign of a chip in trouble, GK102 missing completely from a roadmap, sign of a chip not in trouble, taping out early, waiting on production capacity while not running multiple respins in the meantime.
They aren't any abnormal yields reported, they are lower than usual, that is bog standard on a new process, you would normally expect more of the 7950 parts to both be made, and due to yields be possible than 7970's, we will probably see a 7930 in the not too distant future as a more heavily salvaged part.
There is a reason you VERY rarely, if ever see anything beyond a 2nd salvaged part(5850 has circa 10% of the die fused off, 5830 has about the same again, that's two salvaged parts). If yields are so bad you have to have a third, you basically stop production, likely never get into full scale production and go for a full respin.
Fermi/GF100 was the last and only chip I can remember with 3 salvaged parts, as the 480gtx itself wasn't a full chip, with the 470/465gtx also.
Anything less than 70% of the chip working, and your next performance stop is, 60%, and you'd expect your midrange part to perform somewhere between 50-60% as fast as your high end part, making it pointless and vastly more expensive to make, to sell at the same price as a FAR more profitable midrange part. A 350mm2 chip being sold at £300 is more profitable than a 250mm2 chip at £150, but that same chip is more profitable than selling a 350mm2 chip at £150, which infact would almost certainly be making a loss.