Core overhead

[Yury Kats]

Does anybody have an idea, or better yet, measured data,
to see what the overhead of a core is, both in memory and speed?

For example, what would be the difference between having 1 core
with 100M documents versus having 10 cores with 10M documents?