http://media.hardwareanalysis.com/articles/small/10432.gif" alt="Hardware Analysis - 533 MHz FSB" border=0>
We have the same story here as well. ScienceMark's DSP-like algorithms have proved more dependent on brute-force processing power than on memory bandwidth. As we showed when the 2.40 GHz Pentium 4 was first introduced, the memory subsystem (and therefore FSB bandwidth as well) had virtually no effect on the ScienceMark algorithms. Here we see that pattern played out again, as the average increase here is on the order of 1%, if that, for the increased FSB speed. The clock speed bump to 2.53 GHz produces an increase on the order of 4-5%, as might be expected.

Those of you most interested in scientific, or DSP-like software should look toward a faster processor, perhaps in lieu of a faster bus system.