The clusters previously described use ScientificLinux 5.x and 6.x, and OpenSUSE 12.2 respectively. The Intel compilers and libraries are kept up to date and the OpenSUSE machine benefits from a recent version of gcc/gfortran. MKL's optimised BLAS and LAPACK routines are also available along with precompiled PBLAS/ScaLAPACK/BLACS libraries. The MPI implementation available is OpenMPI 1.6.x with 1.7.1 (testing) only used for the GPU code since it is the only implementation known so far to accept pointers to divide memory and utilise the driver interface for direct memory access. This feature reduces the latency of transfers and is also significant for one part of the code.