Soleron on 29 January 2013
@BlueFalcon
Well the problem is that GPU vendors have hit the efficiency per transistor limit, i.e. what can be done with a GPU's design, so that we'll only get 5% here and 5% there from now on with small improvements rather than huge leaps.
This is combined with the problem that the industry has been and will be stuck on 28nm for a while, because process development is too expensive to fit a half-node every year cadence like 110->90->80->65->55->40 was.
It's incredible that five years after a $90 HD4850 we're only just now seeing equivalent value.
