1 more thing (just found out),
on Xeon E5-1650 processor, use 6 threads for best result.
Tested with 12, 11, 10, 9, 8, 7 = none gives more then with 6 threads.
With this processor your minning goes ~290-310 @6 threads.
of course because your processor has "only" 12MB L3 cache. You need 2MB for one thread.