Dear users and programmers,
we have setup several (1776) ocular motor DCM models (fMRI study) with 8 different areas in SPM8. If we use Version 3684 from January 2010 our estimation time for a single model is around 10-18 Minutes independent of the system/computer used (Windows XP, Windows 7, Ubuntu 10.1, Snow Leopard, Matlab 2009b or 2010a, i7 or Xeon or Core2Duo processors).
But if we upgrade to the latest Version of SPM8 (4010) the estimation time increases by 16-25 fold (160 - 250 minutes!!) and uses up all available threads on a multicore machine whereas the January version only uses one thread. This makes the analysis even on the most recent hardware for a group study (21 subjects) very impractical without a personal Xeon cluster equipped with Matlab licences at your hand. Does anybody know if this is a known side effect and a tribute to the more sophisticated possibilities with the new DCM10 framework or maybe a bug?
Best regards,
P. zu Eulenburg
|