urs truly rant: unified solver in xsi - a possible reality

[Strang]

i would argue that physics in games engines aren't precise enough for what film/animation would need. so you simplifying it to..."... all softimage needs is to probably license and integrate this properly on top of the current simulation framework..." doesn't mean it is that easy. i think people tend to get a wrong impression about the developers of the software we use. they are a small team! that have to work on a large piece of software and have to keep a large group of people happy with new features and maintain the software on all levels.

i used to love to speculate but i would rather get the answer from softimage?

[oktawu]

i totally agree with u steven, on the fact that behind these tools are small teams of people which need to satisfy the need of a larger mass, also, this is indeed just speculation, all im doing is laying my thoughts down. but, because softimage is a small team, it would make much more sense to integrate a dynamics engine rather than building one from scratch. lets look at the current implementation of the psysX rbd engine inside xsi.
they had ode which was a clear example of an api which was not suitable nor accurate enough. lacking the resources or probably not willing to invest too much in the development of something which already existed, they went out and licenced the ageia rbd engine. u say realtime dynamic engines are not accurate enough. i would beg the difference. xsi's rbd engine is more than accurate, and it is based as we all know on the realtime ageia sdk. same can be said about their implementation of cloth, fluids and deformable surfaces. why would a part of a larger api be less accurate than the others, using the same solver. that just doesnt make sense. same can be said about max's havok system. it is based on a realtime api, yet it has served guys like blur, rezn8 or cafefx very well, and still does a great job, for realtime or film quality output.
the argument just isnt solid enough.

[Strang]

i was wrong, game engine physics engines can certainly be used. the accuracy needs to be stepped up though.

ode isn't terrible, it was used on ratatouille. cited here... http://graphics.pixar.com/AnyoneCanCook/paper.pdf

i currently work at blur studio and have worked at cafefx. not sure the parallel you are trying to draw. but i know the tp's ageia integration isn't getting raved reviews. sooo dont believe the hype?

[oktawu]

yeah, im aware of ur whereabouts...its a small world.
my apologies. i was kinda vague with that argument. all i was trying to imply was that numerous companies have used realtime api's for dynamics with huge success (this case havok), both for film and game outputs.
i had no idea ode was in use at pixar ,but as with everything coming from pixar, its probably just the base code they are using, and they have their own implementation of it, since its open source.
i havent heard anything good nor bad about ageia, and these are just ideas.
but looking at the nucleus system autodesk is developing, and looking at the way it currently handles fluid simulations, its pretty clear they are taking the same route as everyone else. using rigid sims on dense metaball surfaces to simulate a fluid like behavior. nothing to do with cfd's. same approach ageia took some time ago.

[Strang]

so nvidia buys ageia... lets speculate some more :)

http://www.engadget.com/2008/02/04/nvidia-to-acquire-ageia/

[visualboo]

Hehehe nice. Nvidia is seriously positioning itself in a great place right now.

[mocaw]

So next they buy sysflex and have all 3 of the "third party" parts of XSI?

[dwigfor]

So next they buy sysflex and have all 3 of the "third party" parts of XSI?

And then Autodesk will buy them......

[janimatic]

i wonder why noone started to write some sample wrapper xsi compiled operator for ageia.
It looks like an easy to use, elegant sdk. see :
http://irrlicht.sourceforge.net/phpBB2/viewtopic.php?t=17669
The problem for a basic op is that it would loose modularity, connectivity, which is the power of this "unified" framework.
So separate operators with io ports should be made, but then that becomes as complex as writing moon dusts.
We need a standard highlevel API that defines connectivity .

[mocaw]

And then Autodesk will buy them......

No worries there- last time I checked Nvidia was worth more than Autodesk. Things could change...but Nvidia would have to sell it off, lose a lot, or Autodesk gain even more. Granted it's a lot more complicated than this so who knows...but if they wanted Nvidia in whole it wouldn't be a drop in the bucket like many of their other acquisitions.

Kind of shows what a relatively small world CG still is in terms of profits/size of companies!

http://quotes.nasdaq.com/asp/SummaryQuote.asp?symbol=NVDA&selected=NVDA

http://quotes.nasdaq.com/asp/summaryquote.asp?symbol=ADSK%60&selected=ADSK%60

For comparisons sake here is apple and adobe:

http://quotes.nasdaq.com/asp/summaryquote.asp?symbol=AAPL%60&selected=AAPL%60

http://quotes.nasdaq.com/asp/summaryquote.asp?symbol=ADBE%60&selected=ADBE%60

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.