Better 'bench' but significantly slower performance

I'm running a Monte Carlo-based simulation requiring multiple iterations, which is running notably slower (taking around twice as long) on my better spec laptop (i7/8GB/2.4GHz) than it is on my others (i5/4GB/1.9-2.5Ghz and i3/4GB/2.16-GHz). While parfor boosts the speed on the i7 (which is a quad core, compared to the dual core of the other two), the fact that at a baseline it is running significantly slower than the others is concerning. All are running 2015b 64-bit, changing to 2016a made no difference (and trying a 32-bit version didn't result in an improvement). The i7 'bench' results indicates that it should provide faster performance than the others.

Are there any possible causes of this, and potential solutions?

[Walter Roberson]

R2016a is not out yet. I had not heard that it was even in beta -- beta usually starts in the new year. Furthermore there is no 32 bit R2016a for any platform.

[the cyclist]

R2016a prerelease is available. I believe it would violate the user agreement to comment on it in any way.

[Anoush Najarian]

Could you please tell us more about the laptops? Certainly the number of cores is different across the machines. Further details on the CPUs, OSs, any VMs, other software running alongside MATLAB can help! Have you tried running any diagnostic tools on the 'slow' machine?

If you are running Windows, please look at http://www.thewindowsclub.com/processor-power-management-windows

One could try profiling the 'slow' i7 and compare to the faster box to see if any specific code is running more slowly.