There's nothing to salvage if they have < 10% yields. They would make far more money selling GT200 than they would GT300. It wouldn't be any different to what ATI did with RV670. Give up the high-end and do what you can to survive until things turn around.
I meant salvaging your entire development. once you've taped out it's obviously not smart to call it a day afterwards. These "tests" with a limited number of wafers running a fast-track are not adding the most costs ofcourse.
AMD could've killed R600 when they knew about their leakage problems, but they didn't maintaining some kind of face/dignity by slotting it lower than it's intended competition.
With GT300 however, what is it's intended competition? if it's LRB, it's still in a hot-seat to be out of the gates on time, if and only if they manage to solve their problematic "power" issues.
So now, the "sudden" (ahem) resurrection of GT212 makes more sense, not an ambitious 384-bit design, but it looks like someone or something smacked some sense into the engineers, maybe it was the wall they were running into?