This gigantic jump in efficiency is just from high mem speed/QPI link?