With 100 single 1Gev mu-, Bogus stand-alone, only multiple scattering present,
I get the following for the SVT fast model (and fast cylinders for the
rest of them):
  Number of events processed : 100
  User=3.57s Real=15.15s Sys=0.29s
number of events = 100 User=14.67s Real=26.54s Sys=0.36s

  Assuming it's the system time that's of interest I get 0.036 s/event.