Hi everyone,
Houdini is slowly becoming my go-to 3D package, and I'm putting together a new workstation with that in mind. Doing some research, I keep reading posts and articles about how to optimize Houdini for multithreading, and I'm a little confused - isn't Houdini optimized for this out of the box? Of course CPU rendering is multithreaded, but from what I can tell, my entire 4 (!) cores are pretty much all grinding away when doing any type of simulation, be it RBD, cloth, POPs, and anything really. Am I missing something here? If I get a CPU with a lot of cores would I still need to spend time optimizing the software to utilize them efficiently? If so, what does that entail exactly (coding? Clever node management? Other workflow adjustments?)
Thanks for your input!
Multithreading Optimization
4518 2 2- 4mylgl6
- Member
- 8 posts
- Joined: April 2019
- Offline
- Sadjad Rabiee
- Member
- 1391 posts
- Joined: Dec. 2010
- Offline
Hi
Yes, Houdini is optimized for multithreading in out of the box.
Most of the nodes are multithreaded.
In some of my tests even without any optimization Houdini was so faster than Maya on a pretty same setup.
But for advanced jobs, if you need more speed, Houdini let you to do some extra optimization .
E.g you can use compile sop inside of Geometry network to compute a group of nodes on different threads or writing OpenCL code to use GPU cores.
Just you should know CPU with more cores is not the only key factor to speed up your simulation.
You should thinking about good GPU too as in last versions of Houdini most of the nodes and solvers are using GPU too.(OpenCL)
You would experience a huge difference when using OpenCL specially in dynamic simulations.
Maybe in cloe future, Mantra will using it aswell.
Yes, Houdini is optimized for multithreading in out of the box.
Most of the nodes are multithreaded.
In some of my tests even without any optimization Houdini was so faster than Maya on a pretty same setup.
But for advanced jobs, if you need more speed, Houdini let you to do some extra optimization .
E.g you can use compile sop inside of Geometry network to compute a group of nodes on different threads or writing OpenCL code to use GPU cores.
Just you should know CPU with more cores is not the only key factor to speed up your simulation.
You should thinking about good GPU too as in last versions of Houdini most of the nodes and solvers are using GPU too.(OpenCL)
You would experience a huge difference when using OpenCL specially in dynamic simulations.
Maybe in cloe future, Mantra will using it aswell.
- Sadjad Rabiee
- Member
- 1391 posts
- Joined: Dec. 2010
- Offline
-
- Quick Links