Usually I wouldn't reply to a Jon Doe, but here goes.
See this thread:
You will see clockspeed is most important, dual CPU will not make a simulation run much faster. Well actually it will make simulation run faster, but you will probably choose a lower clockspeed, which makes dual CPU a bad choice. The sweet spot is probably @ six cores. It might also be the way the study is setup i.e. try to avoid solid mesh and try different solvers to see which one is best for the simulation type you run.
Thank you for replying Richard There are Xeons with more cores. Are these better or are i7s even better? You say that the most important thing is clockspeed so Xeons are not on par with i7s in that level. Would watercooling help this out? Thank you again Richard and I will change my name
clockspeed is key (water cooling would allow you to overclock and get more clockspeed likely)
ssd is key
keeping your problem in ram is key
but in the long run, setting up the problem effectively is the best way to get good solve time. i recently worked with a customer that had 9hr solve times, we got it down to 2hrs with similar accuracy just by changing the way that they setup the problem. until then, they were just throwing hardware at the problem. email@example.com if you're interested in looking into this option.
note regarding cores, at some point the number of cores vs solve time flattens out. i think around 6-8. so unless those are fast cores, i'd still go for lower cores higher clockspeed. and remember, no hyperthreading.