What hardware makes simulations bake faster?


I want to upgrade my system so I can bake cloth simulations faster, but I don’t know which computer component affects simulation speed. I hope someone could tell me what component upgrade would have the most impact on performance of cloth simulations.

My current specs:
Processor: AMD FX 4100 Quad-Core 3.6 GHz
OS: 64 bit Windows 7
Graphics: Nvidia GT 610

Thanks :slight_smile:

From what i have read, simulations are all about CPU power.
But i have no idea if the simulations in Blender are multi-threaded yet, i remember someone asked last year and it was only single threaded at the time (and so not fully exploiting the power of having multi-core/cpu).

For multithreaded support for simulations you need an OpenMP build of blender

Thank you for your responses.

So Richard Marklew, I take it that 2.66a isn’t openmp? In that case, can I find one at Graphicall that still has features from 2.66?

Like this one for example http://www.miikahweb.com/en/blender

Thanks Richard, I really appreciate this! I’m starting the download now.

So, anyone know when this feature will become standard in official releases?

Just test it out, works beautifully XD
Nearly 4 times faster bake times!!!

Thank you Richard and Sanctuary, you two helped me out alot!

Doesn’t the official release use OpenMP? I tried the Miikah’s build but didn’t notice any difference in rigidbody, fluid or smoke simulations compared to the 2.66 release.

In what situation did you get the 4x speed up?

Official stable 2.66 Blender should use OpenMP , in the Blender top menu click on Help then on System Info
Then change the viewport to the Text Editor and open the system_info.txt from the Text Editor browser :

Look now at the build cflags and build cflags and you should see the openmp flags :

I would then be very interested to find the reason of why there is a 4 times speed up, what is special with the MikaHweb builds ?
Could it be that instead of special compile flag there was a bug in stable 2.66a regarding the baking time that has been fixed in svn since (and so the recent build at mikahweb benefit from it ?)
It’s a possibility, as bugfixes are going in svn daily, by example on a recent buildbot (compilled with the same flags as official releases) i noticed when doing a Blender Internal render that it’s faster than with the stable 2.66a

I’m not sure, perhaps it was a random memory leak that got patched, I don’t know.

I also notice that BI was also faster when rendering. I’m not quite verse in the mechanics of this, so I can’t even propose a proper theory to explain the phenomenon :frowning: