True. Although you have to differentiate between those sites that benchmark hot and cold or set a fixed boost for Kepler (PCGH and HT4U do this).

It has become quite complicated. Fact of the matter is, just like with AMDs uber mode, Nvidias cards can clock at or near their maximum boost bin (the exact value depends on each single card, but the variance isn't that large) if comparable settings (temperature target) are applied. Also fact is that most sites test AMD uber vs NV stock which under realistic conditions (pre-heating) is biased towards AMD, not Nvidia. Even Heinz should see that.