aggregate cpu clock speed vs. cores

[grue1970]

i know that single threaded apps benefit from higher clock speeds.  i guess the trick is to find balance between app performance and rendering performance...

so a single 6 core cpu overclocked giving 5ghz vs. dual 6 core cpus overclocked to 3.8ghz.  the aggregate is 45.6ghz vs 30ghz.

seems to me, the dual 6 core is going to give better cpu rendering performance.

(both cpu's would be hyperthreaded, so 12 buckets rendering vs 24 buckets).

any thoughts?

thanks,

[dgorsman · Accepted solution]

Clock speeds don't add like that.  6 cores, each operating at 5 GHz vs. 12 cores each operating at 3.8 GHz (or roughly 0.75 the speed).  Given a set number of operations that can run more or less independantly the second will take slightly longer per operation, but will run twice as many simultaneously.  While the first is partway through its second set, the second is already finished the equivalent number.  Hence the net gain in speed.  Kind of like those grade school math problems "A train leaves ____ travelling at ____, while a car leaves travelling at ___, blahblahblah".

Of course not everything is multi-threaded so you may not be taking full advantage of such a set-up.  And you need to ensure there is sufficient RAM to handle all threads being run; with the extra processor and extra RAM that may be more money than you want to spend.