1) What actually multithreaded in DOP and POP?

2) When i use RBDobject - collisions are multithreaded, but when I use FractureRBD - it's collisions not multithreaded. Is this a BUG?

3) Particles collisions whith geo also not multithreaded?

Some DOP nodes is multitreaded, but some - is not.. This is not a bug. When you work with RBD in other software, you work with one program module. The one “closed” program code block - is more simple for multitreading code optimizations, or for parallel computing. In “closed” RBD scheme developers can use many “fake”, or dirt computations. For example, dirt velocity or position predictions. Yes, this fast. But. You can not change anything inside this module.

I think, Houdini have a best workflow with file-cache-data. You can create RBD for big chunks, drop to file, and then run 2-3 processes with this file-data for computing the small geometry dynamics, for example

Anyway, i hope, SESI can find the some nice way for code optimization in DOP-s and Fluids*. It very needed, because buying 4 or 8-CPU-s computer, not a big problem.
This look some strange, when you have 4 or 8 cores, and you sit, wait, and only one CPU is work

I like houdini dynamic and caching workflow, but sometimes needed dirt and fast simulations :-)
The new XSI 7 also have node based workflow for dynamic. It multithreaded and much faster than in Houdini.

aliax
I like houdini dynamic and caching workflow, but sometimes needed dirt and fast simulations :-)
The new XSI 7 also have node based workflow for dynamic. It multithreaded and much faster than in Houdini.

I`m have no skills with XSI using. For dirt and fast particles simulations, you can tune many DOP-s options, like collision substeps, or collision geometry resolutions.

Of course, if from-package-to-package-and-back jumping is the choice - may be you right