I will keep it simple: We are given 2 CPU: same architecture, same cache per core, same frequency but one difference, one is a dual core one is a quad core.
In path calculation quad will be faster?
I'm asking this because out of curiosity I've checked with NX 7.5 and the results were something like this: on a Xeon dual-core the load is ~50%(1 core), on a Phenom 2 triple core load is roughly ~35%(1 core again) and on a i5 quad the load is, yeap, ~25%.
If you use the traditional "Generate" on a single operation, usually only one core is used. There are places where multi-threading happens, and we are expanding this in to bottleneck areas, so we will take more advantage of multiple cores in the future.
However, if you use "Parallel Generate", then all available cores (up to the limit you specify) are used. Even for a single operation, the processing will happen on another core, while you continue to work in your session on the current core. If you parallel generate multiple operations, they can all go to one core, or be distributed to several cores, depending on how you set your customer defaults, and if there are IPW dependencies that require sequential generation.