Lagoa Multiphysics Engine 1.0

[syzygy]

Beautiful stuff. Will certainly figure into my upgrade plans once cards incorporate this new math.

Click ---> http://gizmodo.com/5591613/incredible-physics-engine-yields-some-seriously-jaw dropping-dirt

 

[MrK6]

Pretty damn awesome stuff, I'd say, thanks for posting. Hopefully it runs smoothly enough to be incorporated into games well.

 

[ModestGamer]

sweet!

 

[jvroig]

I hope in 2 or 3 years, in-game physics can reach levels of sophistication as that Lagoa demo (the game I last heard that has the most realistic physics is Mafia II, and it's not quite there yet, probably limited by currently available hardware). I suppose in 2 or 3 years we would have the necessary CPU and GPU power for it in games. Here's to hoping.

 

[evolucion8]

That was impressive, finally a phisics engine that simulates the water like real water, not like some sort of whittish gel/oil thingy or wax (Like in Cryostasis).

 

[Skurge]

that gravel has pretty low friction.

 

[Bill Brasky]

"Lagoa, like thiago’s previous addon SPH fluids (which this will replace at some point), is for the nodal based programming language ICE that has been in softimages XSI (now autodesk softimage) since version 7, geared mostly around particles and physics, Only the simplest of examples are real time since that is not the objective of this software, being mainly for precalculated effects for use in tv, film, game cutscenes etc etc.. very exciting stuff for 3d geeks since it could eventually mean more sandman type effects from spidy 3 without having a giant team working away in houdini for a year or 2"

 

[at80eighty]

excellent find syzygy - thanks!

 

[syzygy]

More info. The designer is apparently tied to a desk at Ubisoft. He is
a happy man who can't wait to see his baby to market.

The Lagoa Multiphysics 1.0 engine has been developed by Thiago Costa, who currently works as Lead Technical Director at Ubisoft Digital Arts in Montreal. He’s also worked on a smoke simulator which can be seen in the video below:

Link to Thiago's smoke simulator --> http://www.geek.com/articles/games/lagoa-multiphysics-engine-is-simply-stunning-20100720/

 

[nyker96]

that look pretty amazing, I wonder what hardware requirement it has? And why is Ubi doing this not NV or ATI? great stuff, hope to see it in game soon.

 

[gorobei]

good grief charlie brown

Do you people read the link articles? For those who didnt read Zaitsev's post:

THIS IS NOT AN ENGINE. This is a plug-in for an EXISTING Professional 3d Animation Program (Softimage's XSI) with its own Native Physics solver. It allows for more control of the particle sim parameters.

The DEMO scenes are all pre-rendered/calculated by the CPU. Therefore NOT REAL TIME and NOT SUITABLE for GAMES anytime in the next few years.

Hell, even the video lists the renderer as ARNOLD. An un-released GI render plug-in for professional 3d packages. Even if Thiago got something close to work in a game engine, it would never even look 1% as good as the demo because the lighting would still be Gouraud shading crap. Good luck trying to ambient occlusion shade that on a fermi, your fps would drop single digit frames per minute.

 

[NumericalMethods]

DEM (Discrete element modelling) and SPH (Smooth particle hydrodynamics) are two of the most exciting areas of computational physics for me. These simulations look to be pretty close to the cutting edge so I wouldn't expect that level of sophistication in games for some time yet. The fact that people are using these techniques for game based physics engines is awesome.

nVidia have been using SPH for their water effects since the conception of Physx. That's the first time I have seen DEM in a game engine though. And yes low shear strength (low friction dirt) is one of the current challenges facing DEM, but you have to compromise somewhere to get realistic behaviour.

I can't wait to get into this area at some point in my career (hopefully).

 

[SunnyD]

Wow, their cloth mechanics and especially their elastics/sticky bodies are incredible.

 

[MrK6]

good grief charlie brown

Do you people read the link articles? For those who didnt read Zaitsev's post:

THIS IS NOT AN ENGINE. This is a plug-in for an EXISTING Professional 3d Animation Program (Softimage's XSI) with its own Native Physics solver. It allows for more control of the particle sim parameters.

The DEMO scenes are all pre-rendered/calculated by the CPU. Therefore NOT REAL TIME and NOT SUITABLE for GAMES anytime in the next few years.

Hell, even the video lists the renderer as ARNOLD. An un-released GI render plug-in for professional 3d packages. Even if Thiago got something close to work in a game engine, it would never even look 1% as good as the demo because the lighting would still be Gouraud shading crap. Good luck trying to ambient occlusion shade that on a fermi, your fps would drop single digit frames per minute.

And the destruction effects like what we see in BF:BC2 now were only in CGI five years ago. Everything takes time, and this is looking at the future.

 

[gorobei]

the destruction effects in BFBC2 are pre cut geometry with dynamics applied. Source Engine had that 6 years ago, BC2 just scales up the amount of stuff being sim'd as would be expected with hardware performance increases over time.

A full physics solver like XSI's with additional behavior plugins like lagoa costs major resources. I work in 3d cgi, stuff we did 6 years ago at 30hrs/frame can be done in 1hour/frame today, but that still is nowhere near real time. It will probably be another 6 years before you can do the same effects at 30fps. The time scale of these things is not that fast.

There is impressive stuff being done at scale, look up the ghostbusters demos and the cancelled project offset. But stop equating any "ooh and ahh" demo with next gen, there is stuff I have seen at siggraph that was more impressive but has never even come close to reaching market in 10 years.

 

[jvroig]

There is impressive stuff being done at scale, look up the ghostbusters demos and the cancelled project offset. But stop equating any "ooh and ahh" demo with next gen, there is stuff I have seen at siggraph that was more impressive but has never even come close to reaching market in 10 years.

That's rather disappointing, of course, but thanks for setting it straight. It's nice to hear it as it is from someone in the industry without the hype. Makes the expectations more in line with reality.

A full physics solver like XSI's with additional behavior plugins like lagoa costs major resources. I work in 3d cgi, stuff we did 6 years ago at 30hrs/frame can be done in 1hour/frame today, but that still is nowhere near real time. It will probably be another 6 years before you can do the same effects at 30fps. The time scale of these things is not that fast.

Is this even with Fermi-level hardware? 30hr/frame -> 1hr/frame? Nothing more? Or are you actually using more powerful hardware than top-bin Fermi?

 

[MrK6]

the destruction effects in BFBC2 are pre cut geometry with dynamics applied. Source Engine had that 6 years ago, BC2 just scales up the amount of stuff being sim'd as would be expected with hardware performance increases over time.

Right, and what games actually used it in real-time to the scale that it's seen in BF:BC2 (i.e. much of the environment)? Again, only CGI/prerendered stuff did that.

A full physics solver like XSI's with additional behavior plugins like lagoa costs major resources. I work in 3d cgi, stuff we did 6 years ago at 30hrs/frame can be done in 1hour/frame today, but that still is nowhere near real time. It will probably be another 6 years before you can do the same effects at 30fps. The time scale of these things is not that fast.

That's why I specifically stated "five years ago," with the direct correlation being that five years from now, we might see something to this effect in real-time in actual games.

There is impressive stuff being done at scale, look up the ghostbusters demos and the cancelled project offset. But stop equating any "ooh and ahh" demo with next gen, there is stuff I have seen at siggraph that was more impressive but has never even come close to reaching market in 10 years.

Just because something has the possibility to be "next gen" doesn't mean it actually will be next gen. Would it make you happier if everyone posted "meh." or a mod locked the thread? I just don't understand some people's posts.

 

[gorobei]

Is this even with Fermi-level hardware? 30hr/frame -> 1hr/frame? Nothing more? Or are you actually using more powerful hardware than top-bin Fermi?

I'm referring in this case to render farm time. The farm procs we had back then (pentium4 era) would take upwards of 24hrs to complete a frame, and 100+ procs to complete a 3 second shot in time to be reviewed in dailies. Modern core2 and coreI farms can render that same frame on a single proc in about an hour.

Dont know about fermi render rates, assuming it is optimized for the renderer, there could be some order of magnitude gains, but I don't think that there are that many Fermi farms up yet so there is no data for ballparking.

My point about Fermi was that if you shift the resources over to the cuda functions to calculate just the ambient occlusion for all the particles/sprites in the sand/charcoal shots in the demo, you wouldn't have enough power to render the regular raster components in real time. Self shadow, cast shadow, cubic sampling just take too much render power. That's on the Arnold render side. Fermi might be able to crank out the basic motion collision stuff, but the shear number of objects is still too big.

 

[gorobei]

Right, and what games actually used it in real-time to the scale that it's seen in BF:BC2 (i.e. much of the environment)? Again, only CGI/prerendered stuff did that.

you might equate what is done in BC2 with feature film, but it technically is well short of a full effect.
A full destruction effect would:
1) precut the geometry.(if you wanted a custom look, there are automated plugins that will cut the mesh in a less than esthetically pleasing way)
2) do basic dynamics on those chunks.
3) emit particles from the newly exposed surfaces and run dynamics on them.
4) light and shadow those chunks and sub particles.
5) emit dust cloud particle/sprites to cover the transition animation.
6) composite the scene for best impact
7) throw gaussian blur and film grain on top to cover up any remaining ugliness.

BC2 does steps 2 and 5. Source did step 2 and sometimes 5. Source did maybe 5 fragments and BC does maybe 70. This is in line with hardware improvements. Game engines have been doing this stuff for a while, there just doing more of it. Until they add a few more steps it isn't 'feature quality' to a discriminating eye.

That's why I specifically stated "five years ago," with the direct correlation being that five years from now, we might see something to this effect in real-time in actual games.

And my point was that it will likely be 12 years till we see the effects I was working on come to games. The majority of effects: from bump maps, displacement maps, subdivision surfaces, depth of field, etc have taken on average about 10+ years to transition to games from film use(we have been doing this stuff for a while now). The adoption interval is improving but as we've gone along the screen resolutions and color depth have scaled as well slowing actual in game usage. 1080p resolution, 32bit color, 120hz 3d have and will slow down adoption as much as 2k/4k res and HDRI anchored film effects. The only reason people think this stuff is recent is because pixar has been showing the making of stuff on the bonus dvd. The same pixar that was using photoshop3 back in 2000 when everyone was on ps6.

Just because something has the possibility to be "next gen" doesn't mean it actually will be next gen. Would it make you happier if everyone posted "meh." or a mod locked the thread? I just don't understand some people's posts.

More than a few posters in this thread have conflated this demo with "next gen game engine" instead of reading zaitsevs post about it being a film engine "plugin" because of the fact that the guy is currently working at ubisoft. If any of the early posters had bothered to read the credits at the end of the video or read the comments on the linked pages they would have known this was not game related or real time.
I can either post correct information, or post nothing and let people run around with their heads full of fairydust like tweakboy/bot.

 

[MrK6]

you might equate what is done in BC2 with feature film, but it technically is well short of a full effect.
A full destruction effect would:
1) precut the geometry.(if you wanted a custom look, there are automated plugins that will cut the mesh in a less than esthetically pleasing way)
2) do basic dynamics on those chunks.
3) emit particles from the newly exposed surfaces and run dynamics on them.
4) light and shadow those chunks and sub particles.
5) emit dust cloud particle/sprites to cover the transition animation.
6) composite the scene for best impact
7) throw gaussian blur and film grain on top to cover up any remaining ugliness.

BC2 does steps 2 and 5. Source did step 2 and sometimes 5. Source did maybe 5 fragments and BC does maybe 70. This is in line with hardware improvements. Game engines have been doing this stuff for a while, there just doing more of it. Until they add a few more steps it isn't 'feature quality' to a discriminating eye.

Right, I wasn't proposing that it was CGI quality, just this level of detail was only seen in pre-rendered stuff, and now it's real-time. I'm sure there's tons more we can do today that is simply hardware-limited.

And my point was that it will likely be 12 years till we see the effects I was working on come to games. The majority of effects: from bump maps, displacement maps, subdivision surfaces, depth of field, etc have taken on average about 10+ years to transition to games from film use(we have been doing this stuff for a while now). The adoption interval is improving but as we've gone along the screen resolutions and color depth have scaled as well slowing actual in game usage. 1080p resolution, 32bit color, 120hz 3d have and will slow down adoption as much as 2k/4k res and HDRI anchored film effects. The only reason people think this stuff is recent is because pixar has been showing the making of stuff on the bonus dvd. The same pixar that was using photoshop3 back in 2000 when everyone was on ps6.

Very interesting. Good to hear some "inside the business" insight.

More than a few posters in this thread have conflated this demo with "next gen game engine" instead of reading zaitsevs post about it being a film engine "plugin" because of the fact that the guy is currently working at ubisoft. If any of the early posters had bothered to read the credits at the end of the video or read the comments on the linked pages they would have known this was not game related or real time.
I can either post correct information, or post nothing and let people run around with their heads full of fairydust like tweakboy/bot.

Ah, I see what you meant. Getting excited about hardware and software is cool, spreading misinformation is not, thanks for keeping things straight.

 

[imaheadcase]

Actually BC2 uses predefined stuff. A building falls the same way each time, a tank explodes the same way each time, a jeep has like 3 different ways it gets destroyed.

When i think of "next gen", i'm thinking of when nothing is predefined in script mode. BC2 just makes it seem that way, half life 1 did thinks like that in some degree or the other.