The Case for Higher Computational Density in the Memory-Bound FDTD Method within Multicore Environments