Pretty sure the engineers at Nvidia could build a PCB version of the 590 that would handle 2000W if they were asked to.
After-market modding a 590 so that it can do it is not proof that the original engineers could not.
yes
I suspect the original engineers were given a target production cost, including build time as well as PCB design/development time as well, and the product you see is the best any group of engineers could have come to while maximizing within the predefined constraints.
no
I have no doubt that if Nvidia's engineers were instructed to design the 590 such that it could deliver 50% gross margins while selling at an $800 price-point instead of the current $700 price-point then they would have endeavored to include even more robust PCB components (at the expense of pushing up the BoM for the 590).
no, but close
Overall score 1.5/3 = 50%
You cant build a Ferari using cast iron, but you can build a train, which will move slower but transport raw material in large quantities just as fast. Nvidia tried to build a ferari, because the train was grossly over PCI-E spec... Their competitor who uses more efficient core design, has succeeded at building the ferari.
Nvidia failed, because they could not deliver a functioning frari and they did not deliver the train either.
EDIT:
Just for the reference to above example, TiN is making the train, which 590 should have been.