Hopefully customers will punish them for their poor decisions/designs and AMD will get some much deserved praise for their dual-gpu flagship the 6990 as its VRMs were done right.
590 = 4 phase VRM with 35A mosfets for a max theoretical load of 140A max.
6990 = 6 phase VRM with 80A mosfets for a max theoretical load of 480A max.
Clearly the engineers need a good slapping for this one. I really liked what Phaedrus said on the matter:
It would have been trivial to beef up the VRMs on the GTX590. Replace the 35A DrMOS chips with a pair of 50A mosfets and a driver, beef up the inductor, and they'd need to use a dedicated controller and add their own sense circuitry. Nothing that they don't know how to do, they've done it every time they design a new card. They could probably lift the design from an existing card and just tweak it with higher rated components and some different controller settings. 20hrs work + testing, tops. They could even do it with a Volterra solution to make it even easier and arguably more reliable.
But that would cost about 3x more per phase than the solution they went with--5x for Volterra. At some point the engineers, or the engineers' bosses, said "Ok, we need to make this cheaper, and we need it fast. Let's start over using the cheapest DrMOS solution we can get our hands on."