Here is another nice simulation of fluids
http://www.continuousphysics.com/ftp/pub/test/index.php?dir=&file=Ageia_Airtight.mp4

Quote:Original post by Anonymous Poster
Here is another nice simulation of fluids
http://www.continuousphysics.com/ftp/pub/test/index.php?dir=&file=Ageia_Airtight.mp4

Except there's a big difference - the Meqon video is a real-time capture running on old (existing) hardware. The Ageia video is, well, who knows what it is? There's no reason (that I know of) to think that most of the impressive stuff is real-time, and the bits of the videos that _are_ clearly from an interactive session have the physics running in slow motion (and there's one recent, similar, video that has terrible jitter too), even though it's the kind of stuff that could (just) be run in real-time in software! Novodex liquid running in software is _not_ interactive.

Incidently - I really like the Meqon fracture - very realistic looking. There are some issues with the demo (some jiggling on the ragdoll, jiggling, slowdown and not-sleeping when you pile up lots of tiles together etc) but generally it and the videos are pretty impressive, I think.

I always thought that Smoother Particle Hydrodynamics was a way of cheating the Navier equation of fluids, since it violates the more important characteristic of fluids which is incompressibility.
I have to admit that with good rendering techniques the method can be implemented to convey a very compelling fluids impression. Here is an implementation that is very nice too.

http://www.ss.iij4u.or.jp/~amada/fluid/

Since the method is so easy to implement and it fix rather nicelly with any existing solver techniques and even with ordinary particle systems, my guest is that each commercial engine (havoc and Novodex) will also come up with a more or less very polish implementation that will handle perhaps several thousand particles, specially using the extraordinary brute force power of these new consoles. I also think that the other free engines (True Axis, Newton, Tokamak and maybe even ODE) will follow suit with versions that won’t be as impressive since they do not have access to the new hardware, but with these new dual processors maybe they can do a decent job too.

The future of real time simulation will be very challenging, but that can only be a good thing.

Quote:Original post by jovani
I always thought that Smoother Particle Hydrodynamics was a way of cheating the Navier equation of fluids, since it violates the more important characteristic of fluids which is incompressibility.

If I was being pedantic I'd point out that fluids aren't necessarily incompressible. Air is a fluid, for example. liquids are fluids that are at least nearly incompressible.

Also, you can add incompressibility (e.g. paper by Simon Premoze) to the equations using SPH - my understanding is that SPH is just a way of discretising the equations. However, doing so makes solving the system very slow because of the Poisson equation when you solve for pressure - just like grid methods.

One of the nice things about particle methods is that they conserve volume/mass (well, particles!), which is just as nice a thing as incompressibility, if not nicer - a container full of liquid that slowly evaporates due to numerical error isn't good :)

I like the fracture video. That looks like fun to play with.