In some situations we might limit the number of threads running on a socket or server to less than the total number of physical cores present. (We don't use hyper-threading.) Where we are not using all the physical cores, we could potentially use Turbo Boost. For some jobs, this might lead to better runtimes.
I woudl be very interested to hear from anyone who could share operational experience of this capability; particularly on Dell servers using Sandy Bridge, Ivy Bridge and Haswells