Athlon 64 X2 5200+ dual core on Windows XP. Non-realtime (faster than realtime) app. Fixed time steps - UpdateWorld(0.01f). Each simulation is started with a new world and cache flush. Beta version 18 I think.
Bug 1) When two simulations are run with the same initial settings they should have the same output. When using one thread this is what happens. But when using two threads with NewtonSetThreadsCount(mpWorld,2), there is an occasional lack of determinism (approx 1 in 50 simulations with each simulation having 300 iterations). This can be very minor or more significant.
Bug 2) When using one thread task manager shows CPU utilisation in the high 40s. When using two threads, utilisation is in the high 90s as expected. However, actually simulation speed is unaffected - i.e. it is no quicker using 2 cores than 1.
Thanks, Joe