PhysX Real-Time Fluid Physics Are Crazy Realistic

[dvd_72]

Don't get me wrong, that's really fucking impressive, but it's not quite right. There's something just a touch off about the way the water moves, it seems a bit kinda... wobbly. Now that's just me being pedantic, it's still great.

To be fair, our understanding of fluid dynamics isn't quite complete. I have a friend doing a maths PHD on fluid dynamics, trying to create an equation that predicts how fluids interact, when and where instabilities arise in their flows. I suspect that until we have fluid dynamics completely mapped out our fluid models will look a bit off.

 

[blink]

I can't wait to see this implemented in actual game tidals.

What are you going to stare at the water for anything more than a minute? Maybe it could help in a cutscene or something but really? I thought the water in farcry 3 was pretty impressive when I first saw it but then after approximately half a minute of swimming in it I never noticed the graphics again.

You ever hear the arguments that the videogame industry are going broke because developers are spending too much money on graphics and indie games can't keep up? (I think that's how the arguement goes)

Well things like this are the cause for it.

(sorry for raining on your parade but it seems I'm the first one on this thread who has a different point of view so I'm expressing it strongly)

 

[grigjd3]

Don't get me wrong, that's really fucking impressive, but it's not quite right. There's something just a touch off about the way the water moves, it seems a bit kinda... wobbly. Now that's just me being pedantic, it's still great.

To be fair, our understanding of fluid dynamics isn't quite complete. I have a friend doing a maths PHD on fluid dynamics, trying to create an equation that predicts how fluids interact, when and where instabilities arise in their flows. I suspect that until we have fluid dynamics completely mapped out our fluid models will look a bit off.

I don't know for sure, but my first guess is that this isn't a true fluid dynamics simulation in that they aren't actually calculating anything resembling the Navier-Stokes equation(https://en.wikipedia.org/wiki/Navier-Stokes). From what they reveal, it seems like a particle simulation with independent particle-particle interactions and some constant acceleration due to gravity and a boundary that the particles can't move through. It seems like what's left here is to tweak parameters on the particle-particle interactions to come as close as possible to seeming realistic.

If I am wrong and they are calculating Navier-Stokes, then I am really impressed and I'd love to work with these people.

 

[Vivi22]

You ever hear the arguments that the videogame industry are going broke because developers are spending too much money on graphics and indie games can't keep up? (I think that's how the arguement goes)

Well things like this are the cause for it.

Video game development is expensive, but not because it's particularly difficult to implement simulations using fairly standard graphics or physics libraries like PhysX. If anything, having those simulations already done, ready to go, and built into the library you're using makes things cheaper, not more expensive since companies don't have to develop their own simulations from scratch anymore.

 

[Daaaah Whoosh]

There was a really great 2D game that had amazing water physics that seemed to work like this. I can't remember what it was called, but it involved propulsion by vomiting.

 

[Product Placement]

So it can handle a bathtub load of simulated water. What happens if you try to make an area with a small lake, using this physics engine?

 

[Fluke]

Slightly wondering here what they've actually done since 3 years ago... http://www.youtube.com/watch?v=1JrM4ujLY_A&feature=player_embedded#!

 

[lacktheknack]

It'll be nice if this is less theory and more practice. Tech demos are fun but at the end of the day we don't see a lot of this employed in real time.

(inb4DVSmakesemotion/oceanpun)

That's because it would kill everything at this point.

It'll come along.

For instance, the music of the vid was from Mirror's Edge. Mirror's Edge had PhysX enabled in it, which added completely realistic glass shattering, cloth tearing and wind gusting that wasn't in the console versions. So by extension, we'll see this water become a common occurrence in the next couple years as well.

 

[lacktheknack]

Slightly wondering here what they've actually done since 3 years ago... http://www.youtube.com/watch?v=1JrM4ujLY_A&feature=player_embedded#!

There's subtle differences. The water in your video looks like someone added cornstarch to it, for instance.

 

[PSimbo]

It's nice to see a thread that doesn't immediately descend into vilifying any and all attempts at making games that little bit more pleasing to look at.

For those asking about the maths, what makes this method interesting is that it's not a force-based method ? it's a position-based method. It doesn't use the Navier-Stokes method which is not a particle method and integrates convective forces over a fixed, static grid to get fluid motion (and is therefore pretty much useless for video games) and it's not another variant of the Smoothed Particle Hydrodynamics (SPH) method because it doesn't take the weighted force contributions from it's neighbouring particles (I believe the game Hydrophobia used a very coarse SPH-like simulation). Position-based methods directly estimate the new position at each time step from the previous step's particle velocity and then impose a set of constraints ? a much more stable method when you don't have that many particles. For those who really, really want to know about the maths: [link]http://mmacklin.com/pbf_sig_preprint.pdf[/link]

I don't think there's any danger of this method ever becoming the standard method of rendering water in games. 90% of the time you're not interacting with the water you see. The guy who will really win the cookie in the field of real-time fluid sims is the guy who can mix particle methods with typical mesh water surfaces on an 'as needed' basis.

TL;DR: The future is not the old method or this new method of doing fluids. It's a hybrid.

 

[bobajob]

You know, I have a secondary 400 series card alongside my GTX 680 DEDICATED to that shit. Nobody likes a 50% performance hit, after all. UNACCEPTABLE.
In games that have advanced effects, it is VERY nice to have (graphics whore here....)

It might also help if more than a couple of titles a year implemented PhysX in any meaningful way.
It's not even the consoles' fault(not SO much, anyway, though they're certainly not helping) rather that most people wouldn't be inclined to install a 2nd GPU to sort the physics calculations, or take the massive framerate hit instead.

Come at me, developers.....

If only it was that easy eh?.

 

[dvd_72]

Don't get me wrong, that's really fucking impressive, but it's not quite right. There's something just a touch off about the way the water moves, it seems a bit kinda... wobbly. Now that's just me being pedantic, it's still great.

To be fair, our understanding of fluid dynamics isn't quite complete. I have a friend doing a maths PHD on fluid dynamics, trying to create an equation that predicts how fluids interact, when and where instabilities arise in their flows. I suspect that until we have fluid dynamics completely mapped out our fluid models will look a bit off.

I don't know for sure, but my first guess is that this isn't a true fluid dynamics simulation in that they aren't actually calculating anything resembling the Navier-Stokes equation(https://en.wikipedia.org/wiki/Navier-Stokes). From what they reveal, it seems like a particle simulation with independent particle-particle interactions and some constant acceleration due to gravity and a boundary that the particles can't move through. It seems like what's left here is to tweak parameters on the particle-particle interactions to come as close as possible to seeming realistic.

If I am wrong and they are calculating Navier-Stokes, then I am really impressed and I'd love to work with these people.

You seem to really know your stuff! I wonder what you think would make for a more efficient way to realisticly model fluids, this Navier-Stokes equation, or the method you (and I, as a matter of fact) think they have used in this example? I'd expect the Navier-Stokes equation as, from my position of nearly complete ignorance of the particulars, it seems it doesn't need to keep track of every particle.

 

[lacktheknack]

You know, I have a secondary 400 series card alongside my GTX 680 DEDICATED to that shit. Nobody likes a 50% performance hit, after all. UNACCEPTABLE.
In games that have advanced effects, it is VERY nice to have (graphics whore here....)

It might also help if more than a couple of titles a year implemented PhysX in any meaningful way.
It's not even the consoles' fault(not SO much, anyway, though they're certainly not helping) rather that most people wouldn't be inclined to install a 2nd GPU to sort the physics calculations, or take the massive framerate hit instead.

Come at me, developers.....

If only it was that easy eh?.

I just wait... Eventually, the lovely PhysX effects can be handled by one good GPU, and my lovely 60FPS can be maintained.

It's cheaper that way.

 

[grigjd3]

You seem to really know your stuff! I wonder what you think would make for a more efficient way to realisticly model fluids, this Navier-Stokes equation, or the method you (and I, as a matter of fact) think they have used in this example? I'd expect the Navier-Stokes equation as, from my position of nearly complete ignorance of the particulars, it seems it doesn't need to keep track of every particle.

I worked in numerical relativity as a graduate student and a post-doc which shares a great deal of math with and sometimes even incorporates fluid dynamics. I would guess that the Navier-Stokes equation is less efficient to use simply by the fact that these people decided to do things this way. I mean, I am not completely certain having never met this particular team, but I have met some of the people hired for this kind of work and they generally are well aware of the options available to them for computing strategies.

 

[hawkeye52]

I can't wait to see this implemented in actual game tidals.

What are you going to stare at the water for anything more than a minute? Maybe it could help in a cutscene or something but really? I thought the water in farcry 3 was pretty impressive when I first saw it but then after approximately half a minute of swimming in it I never noticed the graphics again.

You ever hear the arguments that the videogame industry are going broke because developers are spending too much money on graphics and indie games can't keep up? (I think that's how the arguement goes)

Well things like this are the cause for it.

(sorry for raining on your parade but it seems I'm the first one on this thread who has a different point of view so I'm expressing it strongly)

Yeah I can see where you are coming from on this. Looking at some of the videos posted it's impressive that we can produce environments of that calibre but in the end it might be to the overall game's detriment. However I feel that we will soon (When I say soon I mean another 5 years-decade) where we will finally get out of the uncanny valley and hit a graphical peak where developers can say "Well we could develop in a higher fidelity but what is the point other then as a marketing point". Once that happens the rest of the industry will catch up in terms of narrative and gameplay mechanics in an effort to keep up with the rest of the market and also the cost of graphics as a whole will get reduced as Nvidea and AMD will strive to make graphics cards which are more efficient as opposed to being able to output the highest graphical detail.

Though to say that there isn't any indie developers or developers which produce games that have a low graphics but high gameplay values isn't quite true. Just to name a few examples:

FTL
Any number of MOBA's (for me that mainly means League of Legends)
Dwarf fortress (Possibly one of the most complex games around)
Crusader Kings 2 (Or any other game made by Paradox Plaza really)
Mount and Blade
Day of the tentacle
The walking dead game
Any host of the more popular Visual novels (Though really this is more reading an interactive book then playing a game)

One special mention I would like to make goes to Bethesda for being the kings of "OOH THAT LOOKS PRETTY FROM A DISTANCE" and then on closer inspection the textures look like crap but because of this it has some absolutely stunning scenery and is also playable on lower end machines that use graphics cards from 5 years ago that weren't even designed to be top of the line then (GT 210 512MB for example).

In summary I can understand the hesitancy to over focus on graphics and not donate enough time to other parts of the games but like I have said give it little while and the focus will eventually shift off of that and move onto the other, and in my opinion, more important aspects of gaming.

 

[Do4600]

I was going to say that they should make an Aquaman game with this technology but then I remembered that nobody would ever play it.

 

[CyberMachinist]

The things this tech could add too if paired with people with inventive uses for it in games,assuming they can do it.

makes me think of what you could use water for in Bioshock.

still chances of seeing this in games in the near future are pretty slim due to hardware limitations like everyone said before me, and then i get thoughts of people making games where they overdo this tech to compensate for things they are lacking in the game itself or try to use it as a selling point, maybe I'm just being to pessimistic about where publishers will take this.

 

[Codeknight]

It still looks deeply nestled in the uncanny valley of water. Once they get it to look like it's not made of tens of thousands of balls a Jell-O, That's when I'll be impressed.

Little details like realistic water are nice and all, but I'd happily give that up for streamed code and additional bug testing (yes, I know these guys don't make the games).

 

[Something Amyss]

which when you are trying to have a fully immersive experience in a game, is definitely the way to go.

I don't know. I'd say when you're actually immersed in the fluid, the motion probably doesn't need that much attention to physics. >.>

 

[Shia-Neko-Chan]

The only problem with these physics is the water does not have weight, so it does not push anything. That's why it kept getting stuck in the glass when it kept forming closed containers by accident. In real life, the shards would be pushed aside because they're not actually connected and can't keep the water in place, but in the demonstration, the water simply conformed.

That's easily fixed, though.

The only problem is now that PhysX is ahead of OpenCL. if I want to experience realistic water, I have to get an Nvidia card.