RAM & Processor for Simulation

There's also a really nicely done .pdf of studies done by SW over the years with some numerical results in the Knowledge Base under S-018633.  It does a really good job of breaking things down for Direct Sparse and FFE+, different analysis types and how they utilize multi-cores (or don't) and so on.

My interpretation of that .PDF:  You have to be pretty well maxed out on your CPU usage for the extra cores to really make a difference.  It does happen, but not that often and depends on things like solver type, analysis type, stage of the solution, etc.  I think that helps explain the reasons why Jay wrote "Prioritize CPU clock speed over number of cores" in his recommendation at that link.