Peformance of a rule within a cube vs across cubes

[PlanningDev]

Just wondering if there is a performance hit when using a rule that crosses a cube vs a rule that works within the same cube?

The example here might by a driver based planning cube where the drivers could exist in a separate cube vs within the same cube. Just wondering as many times we keep the drivers separate from the output so that you have a clean "Reporting" cube.

[Duncan P]

The significant difference is between ['item'] style references and DB( ... ) references. You have to use DB for inter-cube references but you also have to use them when referencing string cells in the same cube and when the source position in the same cube is variable and dependent on the target, e.g. "previous period" or "hierarchical parent". I have seen a performance difference of 3:1 between DB and [] in some cases (with [] being faster), although I do not know how this ratio varies with the dimensionality.

[PlanningDev]

So the question would be if you took a simple example

A = B x C as your calc

If B and C came from another cube with the same dimensionality using a DB function would there be a performance difference? Under the hood is there overhead to a DB function or is it only self created when you add calculations and dynamic referneces to determine elements within the DB?