For Maximum Likelihood inference:

• GARLI supports multiple cores through OpenMP. It also uses a genetic algorithm approach for parameter optimization.
• HYPHY supports multicore and can perform ML tree inference.
• There also seems to be a PhyML fork with OpenMP suport.

For Bayesian MCMC inference:

• MrBayes supports multicore through MPI (Altekar et al.).
• One can use BEAST on many cores by performing independent runs in parallel and combining the output.

I have just stumbled upon a paper comparing RAxML with FastTree.

FastTree is another fast ML method, they say much faster than RAxML and not so much less accurate... And is available as Multi-threaded FastTree with OpenMP

There's also POY from AMNH that can be run on clusters.

I heard R`s phylogenetic package combined with its multicore package works fine. But I never tried it.

• Fastcodeml http://wiki.isb-sib.ch/selectome/Selectome_requirements