A parallel version of the B2000++ solver for shared memory architecture has been around for some time, although it has never been explicitly announced. Now a parallel distributed version working with MPI and based on the MUMPS parallel sparse direct solver is also available.
Some tests have already been performed at SMR, and the parallel distributed solver is used to solve large solid model problems with a dual CPU Opteron-based machine, each of the CPU's having 8 cores.
Specific speedup tests reveal that the speedup of the current parallel distributed version of the B2000++ solver is relatively reasonable for solid elements (see graph below). The graph below displays speedup curves for a simple solid FE model consisting of 27-node hexahedral isoparametric elements. Such a model exhibits a substantial fill-in of the global factored matrix and is thus a significant test case.