http://gameplanets.blogspot.com/2007/06/physics-simulations.html Soon to be found in a game near you . . . . Riight..
..... i wouldn't doubt it. might need a PhysX card.... or DX10 hardware with the physics processing capabilities of the geometry shaders.... but it'll happen.
I hate to burst your bubble, but those things are definitely prerendered. And it probably took several days of intensive calculations on a beefy $10 million supercomputer! When common household computers will be fast enough to calculate+render those particle/fluid dynamics simulations in realtime, meteorologists will be able to accurately predict weather full 30 days in advance...
Havok demoed a early beta of their engine about a year ago and showed pretty awesome physics being calculated in real time. I wouldn't doubt that this is possible with the current processing power of today, especially in unison with the GPU. http://video.google.com/videoplay?docid=6261074351828388772 http://video.google.com/videoplay?docid=2350042502418485801
Comparing Havoc to the ones posted above is like comparing apples to oranges. I do those things for living. And you can trust me on the fact that you need at least 10 teraflops of computing power to calculate in realtime the fluid dynamics simulations alone. You will need at least 20 if you include realtime rendering of raytracing+photon mapping.
Alec: then i guess you haven't used NVidia's Smoke demo for the 8800 series? rendering smoke in a box isn't the only thing it does. the water effects it can do are pretty nice too.... and they're all controlled by the mouse movement.
I have seen the smoke demo and it's mostly done by simple noise turbulence... it's far from physically accurate.
I realize its prerendered, it's just cool to see where technology has brought that. Personally, I hate fire effects in games, they are always so cheesy.
It will be a while before we see stuff like that in games. Even so, the water simulations don't look too realistic to me.
Havok takes physics simulations and attempts to optimize it by taking shortcuts. You can count on physics simulations inside 3d applications being heavily unoptimized and based on absolute simulation. That isn't necessary in a video game, but I'm pretty positive that a dedicated math card could chew through a more optimized game oriented physics simulation with the same graphical detail easily. As for the Raytracing and Photon Mapping you haven't seen the previews for 3DSMax 10 and the new mental ray have you? You should look at the mental ray CUDA demonstration. A 3.5 hour render with Final Gather, Caustics, GI, and all that other fun stuff enabled on max 9, took 13 minutes on Max 10 with CUDA.
Well Alec is definitely right about the physics calculation part with the current way my original post was worded. Fluid calculations in professional products can be tuned to be extremely accurate, which is probably what was used to make these videos. What I really meant to say in my first post, was that we could produce similar results in games today as much of that accuracy is unnecessary in a video game environment. Even things like photon mapping can be heavily optimized, I bet in that video they were emitting 50k+ photons per light source. In a professional environment that kind of accuracy might be needed, some developer will come along though and find a way to knock that down to 5k photons, make them calculate faster, and fake the rest. Sure the end result may not be as accurate as using a crap ton of photons, but it's going to look just as awesome.
Of course, I agree 100% on the faking part. Everything can be faked, provided the skill of the CGI artist (a.k.a. "The Magician" ) is adequate enough to make the audience believe in the distorted image of "reality". But, since I was commenting the videos in the first post, I got to say they are extremely accurate. I can fake them with a bunch of tricks and make the rendertimes extremely short, but I doubt I could make them as convincing as the true simulations, since in some areas our brain becomes very picky when certain details are involved. By the way, I am extremely interested in MentalRay GPU hardware acceleration. Can you provide us with some links to the actual testing?
I'll see if I can dig them up, there was a thread on it over at cgtalk.net but I believe it was under one of the general sections, not the max one. I'll find it though.
