PDA

View Full Version : Dynamics calculations too slow



AmigaNewTek
05-11-2011, 03:22 AM
Hello everyone,

i'm trying to realize a waterfall using hypervoxel and a particle emitter (5000 max particle), but the calculations are too slow on a quad core. Any tips to improve the speed? There's another way? I've heard that Dstorm liquid pack can use ageya physics, is that true?

Thank you

Regards

3DGFXStudios
05-11-2011, 03:54 AM
Realflow!

kevman3d
05-11-2011, 04:16 AM
Hello everyone,

i'm trying to realize a waterfall using hypervoxel and a particle emitter (5000 max particle), but the calculations are too slow on a quad core. Any tips to improve the speed? There's another way?

I don't tend to see particles being very slow to process - and speed will of course depend on what else you have in your scene as well...

Sometimes over-complex 3D geometry can be a cause of calculation speed problems if its being used as cloth, collision, etc. Collision objects can get away with fewer polys vs. lots of them...

Any clues? Is it just 5000 particles, or is there huge amounts of complex 3D terrain, or a collection of dynamics like cloth/wind/gravity/hardfx/etc that's involved as well?

AmigaNewTek
05-11-2011, 04:21 AM
I don't tend to see particles being very slow to process - and speed will of course depend on what else you have in your scene as well...

Sometimes over-complex 3D geometry can be a cause of calculation speed problems if its being used as cloth, collision, etc. Collision objects can get away with fewer polys vs. lots of them...

Any clues? Is it just 5000 particles, or is there huge amounts of complex 3D terrain, or a collection of dynamics like cloth/wind/gravity/hardfx/etc that's involved as well?

The particle hits a wall created using a subdvided rectangle and fractalized, maybe this cause a slowdown. Moreover, there is such a river where the particles falls. Could be this to slow down the things? Particle self interaction could slow down things too?

I should check.

thank you

Btw, i see that only one processor is used @ 28% max during calculations.
When there will be an improvement in that way?

Dodgy
05-11-2011, 06:38 AM
I've heard that Dstorm liquid pack can use ageya physics, is that true?

Thank you

Regards

That's true. I'm not sure it works in 10 though, you'd have to ask. They even sell an Ageia board in a pack with it (or at least they did).

3DGFXStudios
05-11-2011, 07:18 AM
Buying Dstorm liquid pack wouldn't be such a good idea. It never going to look as nice as realflow stuff. I think even the free blender option is better.

OnlineRender
05-11-2011, 07:33 AM
try and break the dyn up into parts if possible and simulate each part one at a time and save out the motion , only draw being you may need dynamics to interact with each , but you can work around that .

trying to simulate a whole scene in LW will either just hang or go mega slow .

RealFlow is great but a cheaper alternative would be to use blender then re-import them back into LW I dont suppose you bought Messiah :)

Greenlaw
05-11-2011, 08:21 AM
Also, if you haven't already done so, create a separate scene for your calculations and remove everything but the objects used in the simulation (usually this means just the emitter and collision object.) It sounds like you're using a sub-d object as a collision object, so you will want to freeze that by doing a save transformed and replaceing the sub-d object with it (no need to make Lightwave work a lot harder for this when the results will be the same.)

When you're done calculating the sim, save out the .pfx and apply it to your final 'render' scene.

BTW, it's typical to do any dynamics calculations in its own scene and use only the 'baked' data for your final scene. This practice is usually easier to do and the reults are more efficient for rendering.

Hope this helps.

G.

erikals
05-11-2011, 11:54 AM
That's true. I'm not sure it works in 10 though, you'd have to ask. They even sell an Ageia board in a pack with it (or at least they did).

i mailed them some time back, (bout 6 weeks ago maybe) unfortunately it does not work in LW10.

AmigaNewTek
05-12-2011, 01:31 AM
Also, if you haven't already done so, create a separate scene for your calculations and remove everything but the objects used in the simulation (usually this means just the emitter and collision object.) It sounds like you're using a sub-d object as a collision object, so you will want to freeze that by doing a save transformed and replaceing the sub-d object with it (no need to make Lightwave work a lot harder for this when the results will be the same.)

When you're done calculating the sim, save out the .pfx and apply it to your final 'render' scene.

BTW, it's typical to do any dynamics calculations in its own scene and use only the 'baked' data for your final scene. This practice is usually easier to do and the reults are more efficient for rendering.

Hope this helps.

G.

I will try. If i use a non subd object, a plain block, i wouldn't be able to realistic simulate the water flowing in the object, tho

thank you all for the reply

OnlineRender
05-12-2011, 02:36 AM
generally the best thing to do is create a low poly version of the object and use that then switch it over when you render .