I do have one question that I never resolved in my own mind - how does MAXDOP work with HyperThreaded CPU's?

If, for instance, on a Dual-Xeon/P-IV system, SQL Server thinks it has 4 processors.  Each virtual processor isn't really an entire processor independant of one of the others in its own right, so increasing the load on one will reduce the available load on the 2nd within the pair.

Does SQL Server have the smarts to get around this, by not allocating busy threads to the 2nd virtual CPU in a pair?  Or does it try to squeeze both through?  Are the internal algorithms tuned for Hyperthreading in some way?

In regards to licensing, I'm aware that you don't need extra licenses for these Hyperthreaded chips.