ClustalW-MPI! is the right thing to use if you're interested in using something that's based on ClustalW-1.8, which has been in use (but not under development) for ages.
ClustalW-2 is pretty much just a C++-reimplementation (mainly to make the GUI portable) with the addition of some minor features and some bug-fixes. I don't think anyone has implemented a parallel version of this, but there's not much point in this because of the similarity to version 1.8.
And just to add to this: Clustal-Omega is completely new and at least the pairwise alignment stage is fully parallellized.