Computational physics in material science has made great strides in the past few decades, mainly thanks to the development of faster computers and new algorithms. In particular, the great success of the quantum mechanics based method known as density functional theory (DFT) in its varios flavours has ben responsible for the development an entirely new discipline, aiming at the in silico design of new materials for technological, biological, medical and other applications. However, destpite its wide range of applicability, DFT still lacks high accuracy, and methods to systematically improve it. Quantum Monte Carlo (QMC) methods are promising techniques to overcome the DFT accuracy problem, and although several orders of magnitudes more expensive, they are well suited to exploit the massively parallel supercomputer of today and the of future. However, here I will show that the application of QMC methods in their traditional formulation can result in significant errors, restricting therefore its range of applicability. These errors result from a non size-consistent implementation of the basic algorithms of the method. I will show how a new proposed algorithm cures these problems almost entirely, and result in a one/two orders of magnitude speed up, opening new exciting posiibilities.