PDA

View Full Version : LW grinding to a halt with inertia deformations and extensive nodal motion control



AbstractTech3D
07-03-2012, 04:50 AM
Hello

I have a scene in which I have 10 ribbons animated, with quite a bit of 3 stage hierarchal nodal item motion control to get the animation just so… with variables abstracted out to control null channels, and with the Inertia plugin applied to deform each ribbon.

However, it really grinds to a halt. To the point where the scene can barely be further modified. 417000 polygons in the scene - with the ribbon objects sub-d'd. Every aspect of the interface takes a good 10 seconds to respond. (However it does still render quite quickly).

The slow down seems to occur at 2 points:
- with the multiplication of my extensive item motion nodal setup - even when just applied to nulls (without any geometry)
- and then very much further when inertia is activated on the sub-d'd geometry - even when display sub'd level is only set to 1 (but rendering at 8)

I'm using an i7 3700k cpu + 16GB ram. But a very cheap graphics card.

Could I expect particularly better performance with a better graphics card?
Or am I fundamentally pushing the limits of LW architecture? (which I suspect.).

And / or is the Inertia plugin not sufficient for my purposes? I believe it is possible to construct a custom nodal inertia setup (replacement to the plugin), although probably very tricky - see the comment on the video one week ago: http://www.youtube.com/watch?v=_WV7ApjIsh0
Would a custom nodal Inertia setup be likely to outperform the existing Inertia plugin?

Thanks for your advice!

SplineGod
07-03-2012, 05:43 AM
Perhaps try using a lower rez object and then mdscanning it. Then metalink the higher rez ribbon to that one.

AbstractTech3D
07-03-2012, 03:31 PM
That sounds like a way forward.

Thanks Larry!

SplineGod
07-03-2012, 04:06 PM
Ill be interested to see how well that might work. :)

AbstractTech3D
07-04-2012, 05:51 PM
Ok, the solution I'm testing at the moment is to use IL_SaveTSeq to create a deformed object sequence, and then the free PO_MDDCompiler to create an MDD from such.

Have been wrestling with LW's factory MD Baking size limitations. (The animation is quite long). Also, I think baking needs to happen on high res object since the deformation generated by Inertia is quite different on a high res object to a low res object. (Unless I've not understood the workflow correctly).

Will let you know how that goes.

AbstractTech3D
07-04-2012, 09:23 PM
…seems to work except that PO_MDDCompiler seems to have a 1000 frame limit

SplineGod
07-04-2012, 09:45 PM
Try appling cloth fx to it. Under the file tab select md scan. Once its done save the mdd file.

AbstractTech3D
07-04-2012, 10:01 PM
Unfortunately it crashes Cloth FX scan pretty early on. Just too many points and too much duration I think.

I have just had some success using Object Sequencer applied to my exported sequence of ribbon objects resulting from IL_SaveTSeq.

SplineGod
07-04-2012, 10:05 PM
Using lower rez proxy meshes isnt working? Ive done similar things and then used metalink to use the lower rez mdscanned proxy to drive a higher rez mesh.

AbstractTech3D
07-05-2012, 05:45 AM
Ah ok, after a little more work I seem to have it working by metalinking to a low res object with Cloth FX.

Good advice, thanks Larry!

SplineGod
07-05-2012, 05:58 AM
cool, glad its working :)