Xeon E5-2643 vs E5-2650; cores and cache vs freqency

Would one recommend two 2.0GHz processors with 8 cores and 20MB cache each or two 3.3GHz processors with 4 cores and 10MB cache? I'm running matlab with the signal processing toolbox only. Memory will be the same in either case. Probably will be doing signal processing and working with matricies.

Thanks!

Answer 1

[Walter Roberson]

If you are using the signal processing toolbox without the Parallel Computing Toolbox, then the only use of multiple cores you would get would be for the cases where MATLAB automatically parallelizes the code, such as when you have one of the common mathematical operations being applied to larger vectors (breakpoint is about 8000-ish elements or more for straight forward operations, much smaller for tasks such as eigenvector computations.) To the extent that you end up confined to one core, the 3.3 GHz is likely to be much more efficient than the 2.0 GHz.

Caution: somewhere around R2011b, people started noticing that in 64 bit MATLAB, some mathematical operations were much slower on some processors. i5 and i7 were noted. I thought Xeon was one of the ones listed, but it is possible I was thinking of this AMD thread.

If possible, benchmark before you buy.