What I will say though, is that a fraction of a second is an age for microprocessor(You could fit over a billion instructions in a fraction of a second now), and the first instructions, before the pipeline has filled, are the slowest to execute, so boosting to counteract that and get to a filled pipeline quickly makes sense for speed & efficiency reasons. I think it's disingenuous to say this does nothing, for most general tasks it probably boosts for the whole execution window and for larger tasks it probably boosts until after the pipeline is filled, both would have some measurable impact(With the right tools anyway) , just not a huge one.