This is probably fixed in SVN trunk, but we need confirmation.
Can anyone verify this behavior? I ran this on a 4-core Windows 7 machine with the OpenMx Beta and R 3.1.0. omxDetectCores() gave me 1, instead of 4 or 4-1=3. The task manager showed all four cores active during optimization, but total CPU usage was around 25%=1/4. When I tried to modify the number of threads used by setting the option to use 3 threads, I did get a crash from R.
Mike Neale reported a protection stack overflow to the openmx dev list. Maybe that is why it is crashing on windows?
Please try with 3650.