I could, but it would be awfully boring: the kill-a-watt changed by less than 1% while running (I assume because of heating and therefore increased omhic resistance). I believe the 10% variance is with respect to unit-to-unit variance, not within the same unit. Also, the hashing variance was really low, all three are pretty much locked onto 60.2 +/- 0.6 GH/s. I believe this is because the number of diff1 shares returned to cgminer was very large, leading to small variance.

Cheers