You are using an old version which is no longer supported, but the main problem (although we cannot see you input script) is probably that with your settings a lot of DFT calculations are started at the same time, on that single machine. Each supercell for the dynamical matrix will be 5x5x1 times the original one, which is a very large calculation. Unless you tell the program to only run 1 displacement per node, it will almost certainly run out of memory. In newer versions the default has been updated, but in older version, set the keyword processes_per_displacement in HamiltonianDerivatives/DynamicaMatrix to the number of MPI processes (or number of cores per node, if you run over many nodes) to avoid this.
