I have two BJ's, one with a board marked Rev. 1.1 and one with no markings that I've found (1.0?).  I am running them at 610.  

Even though it cost me a few satoshis, I did the experiment:

The Rev. 1.1 board draws 485 watts at the wall (well, at the surge protection, actually).  The unknown board draws 470.

The Rev. 1.1 board seems to hash just a little faster on average, but we're only talking a handful or less g/hash.

At 600 clock rate, the Rev. 1.1 board draws 474, while the no-label board draws 464.  

Since I don't think there's much of a detectable difference in hashrate between setting the clocks at 600 or 610 (a couple or few g/hash per machine... i.e. about 3 watts per g/hash), and given the risks of overclocking, I've gone down to 600.