My advice: don't run ML algorithms on a laptop. At least not one with a cheap processor (top of the line Core i7 is a different beast I imagine). From my experience, you're better off buying the cheapest laptop and running stuff on a server purchased/rented in part with what you saved. Or maybe renting an EC2 HC instance. Disclaimer: I work at a (ML/NLP/IR) lab so I didn't pay for the servers but it's often the difference between something taking 12h on my 11"2012 MBA, 6h on the slow servers or 6mins on the best one (which just has a better processor than the others).
> 6h on the slow servers or 6mins on the best one (which just has a better processor than the others).
60 times differences? What kind of CPUs do you have in slow and fast servers?
Presumably it's parallelizable. Double the clockspeed, 16x the cores... that'll get you 32x improvement right there, but you've also got way more L3 cache, which might help considerably.
plus much more ram, faster drives and the fact that newer cpus get more done even at comparable clockspeeds ... and there's your 60x