Profiling shows Time spent more than possible
After profiling windows-service (~900 threads) during 117 sec on 16-core server, I've got the one method's CPU time is 52% (1 395 660 ms)
So I count the core number that can provide me such time: 1395/117=~12 core.
If it's correct, 12 core should work whole time to get method's 117 sec.
But it's only 52% according to measurement result. and 100% should require 24 core (instead of my's 16)
So the question is how is it possible? Any thoughts?