First-principles density functional theory (DFT) based, atomistic, self-consistent device simulations are performed for realistically sized Si nanowire field effect transistors (NW FETs) having tens of thousands of atoms. Through mode space transformation, DFT Hamiltonian and overlap matrices are reduced in size from a few thousands to around one hundred. Ultra-efficient quantum-mechanical transport calculations in the non-equilibrium Green's function formalism in a non-orthogonal basis are therefore made possible. The n-type and p-type Si NW FETs are simulated and found to exhibit similar device performance in the nanoscale regime. Published by AIP Publishing