I have two PCs that I routinely use for Flow Simulation (running version 2015 SP0.0 currently). One of them is an older six-core i7-3960X clocked at 4.5GHz, and the other is a newer six core i7-4960x clocked at 4.0GHz. I have noticed lately that the 4960x simulations seem to be taking too long to run. To test this, I set up the same simulation on each machine side-by-side and let them go. Sure enough, the 3960x @ 4.5GHz was averaging 5 seconds per iteration, and the 4960x @ 4.0GHz was closer to 30s per iteration. I tried running the simulation on the 4960x @ 4GHz again, but set the solver to only use 10 cores (instead of the maximum number which is 12 by default), and the simulation is outperforming the faster machine now at ~ 4 seconds per iteration.
Has anyone else seen such behavior? I'm solving using all 12 cores on the 3960x @ 4.5GHz and it works just fine. Maybe there is an issue with parallel processing with the newer processor chips (or motherboard implementation).