Importance of Cache in the performance of the calculation

Hello,

I am trying to build a workstation for specifically Matlab calculations. I will be working with huge matrices maybe even up to 1 M by 1 M. For me it is important to be able to create large matrices rather than speed that I am solving them. So CPU is of second importance. But I am not sure how the cache amount would effect the performance of the CPU in my calculations. I have two options both XEON processors. One with 10 Mb cache and another with 15 Mb. I am aware that frequency and number of cores are variables too but I just want to understand the role of cache in such large calculations.

Thank you

Erdem

Answer 1

[Walter Roberson]

A lot is going to depend upon how the data is being used.

http://stackoverflow.com/questions/8620303/how-many-bytes-does-a-xeon-bring-into-the-cache-per-memory-access

and search through https://en.wikipedia.org/wiki/Basic_Linear_Algebra_Subprograms for "cache"