Bertrand Hechenleitner and Karl Entacher have performed different simulations to examine bad effects of poor RNGs on simulation results. They suggest two RNGs to be used: Mersenne Twister and the Objected-Oriented Random-Number Package of Pierre L’Ecuyer et al. Source available from the page. Home page: http://www3.fh-sbg.ac.at/~bhechenl/research.htm