I think a more reasoned response is called for. It is perfectly possible for a piece of software that uses one part of the hardware to be slowed down by the overhead of using other hardware that the particular software does not employ. I was just playing with some Cuda apps over the weekend with one and two 285 GPUS in my Mac Pro, and the Mac 285 would do over 100,000 calcs in a fixed time on its own, but just over 80,000 when in a pair. Of course the other card was also doing 80,000, so that the total was 160,000, but there was clearly an overhead in just managing two groups. I think it is possible for an app that does not use more than one group of cores to be slowed down by the presence of another group. If something is well written it is unlikely to happen but I do not think this is a stupid question.