Physics Demo Newton large scale stacking.

[rachetandclank3]

Those looks like potatos... i like potatos :P

[ReacticFlyer]

Would look much smoother and faster if you speed up the faking video. -.-

[JerezJulio]

@ReacticFlyer

why do you say the video is fake?

the engine open source, you can download it from google source and reproduce the demo.

[ReacticFlyer]

@JerezJulio I didn't say fake. By "facking" i mean't fucking. Sorry if you found that offensive.

[JerezJulio]

@ReacticFlyer

Oh I see, No I am no affended.

The video is not prerendered, it is a video capture from Frap.

Now that I am making a Max Plugin I, am othe people will be able to make simulated videos with mass physics.

and maybe prerender them :)

[bluetape66]

@ReacticFlyer why is it fake!

[ReacticFlyer]

@bluetape66 FOR THE LAST TIME. By "facking" i mean't "fucking"

[GeekingForJesus]

This calls for GTX 580s in tri SLI, as well as a GTX 580 as the dedicated PhysX card!

[1223isaiah]

what game is this

[RaymondCorrigan]

For some reason I had this voice in my head going "oh no, oh no, oh no, oh no!"

[GoldenClipz]

is this in unity?

[HTIDbopowell]

you gotta lol at the 1fps

[xxl606]

the fact remains that there are still a lot of BALLS

[TrackManiaMon]

no, not high performance. get a better gpu

[Bananadine]

Hey you know what would make it faster? Removing the third dimension, which mostly just makes it harder to design a good video game anyway.

The screen is planar. The mouse moves in a plane; the arrow keys act in a plane. There are some good reasons to use a third dimension in a game, but there are more good reasons not to.

[RavingKunaiX]

@Bananadine

Truth. 2D > 3D

[iNfx1]

Even on a core i7 system, (which is the best CPU available to the consumer market at the moment) the engine would still lag, there is no solution so far to fixing physics at a pre-rendered level, so these companies stick with what they have.

[ilhadosmacacos]

k when i said this pc sucks i was just trollin :P

but i think ageia physx would brutally increase fps...

anyway,do ageia physx work in any 3d application with colisions?!

[iNfx1]

It still has to support it. Ageia only came out officially last year so I don't think this will support it, however it will benefit from being run on a newer system.

[mit871chell]

They use server grade CPUs for these types of calculations and combine them to form a supercomputer, thats what is used for physics calculations where data must be processed at incredible speeds.

[iNfx1]

Yeah, the Xeon equivalent of an i7. Same socket, same speed.

[mit871chell]

That isn't true Xeons have a higher amount of Flops.

[Pyroney]

the pc sucks? im sure your godly computer could render 27000 at once with no lag

[ilhadosmacacos]

Ok thanks *awesome face*

but nvidia physix can do more than this i guess

or maybe you are right, i was just a retarded kid when i aid that :P

[ViolentBiscuits]

Lol people dont realize this guy was getting about 1 fps thats why it was so slow.

[TheKpow]

It's like Crysis on crack

[gleneth00]

speed it up

[Vikinatorn]

He allready did, about 15 times im guessing, because its not 1fps we are showing.

[gleneth00]

but the vid shud be sped up to a realistic speed

[Vikinatorn]

Yea i agree.

[WhiteDragon103]

I think it is possible to achieve, easily, 60 fps in real-time using these many objects on the very same processor used to animate this video.

[Vikinatorn]

U dont even know how powerful the processor that this was created on is.

But yeah, if u would set down the collision quality by a lot, make the hitboxes simpler, and have the right software, then it might be possible.

[WhiteDragon103]

I know that, whatever the processing power, things can be done to improve performance by a very large percent.

What ideas did you have? You mentioned lowering collision quality. I have my own ideas for how to speed up the simulation, but I'm wondering what you were thinking of.

[Vikinatorn]

Well in some 3D applications u can lower it somehow... I dont acctually know if it makes a diffrence in visuals, but it usually speeds it up. Simpler hitboxes makes it faster too. But i dont know too much about that, because i usually dont need to lower quality in this kind of simulations ;D

What ideas are u having?

[johnathenbates]

wow 1 frame per second, it's smokin Brother!

[bv90andy]

I like the FPS written up in the left corner:1,..FPS

[Headcrabb]

It is forced to stay that way so that when they piece it together it is consistent

[cyberdaemon]

Its probably played out in slow motion so we dont see much of the lag on the screen :D I wonder how good the PC has to be in order to run the same thing at least at 40-80 FPS ?

[MouthofSaurons]

You probably mean 60-75 FPS? Because no screen can refresh faster than 75hz (75 fps). If you don't enable VSync, you will see horrible tearing.

[masterchief3k]

Try this using a tesla!

[mat967]

lol tesla is sooo powerful! id like to use one as a GPU for a game using physx but it does not suport it...

[joejoe133]

1fps is fast but apart from that im surprised its stable with that many objects

[AndyGoRAWR]

Very cool, but I don't ever see when you'll need to render that many objects in real time because what's the point of real time rendering if it's at 1-2 fps =\. Amazing though and very cool.

[JerezJulio]

Thanks you.

The difference is that this simulation is realistic and run at iterative time.

That level of realistic simulation is not possible with other real time simulation tecnology.

This a simple core2, but if you extend it to 4or 8 cores the it works in pure real time which is a blessing for people rendering large scale animations like SketchupPhysics where they make entires cities.

BTW the physics part of this is now a lot faster than you see in the video using the same hardware.

[artisanfx]

woow

[darkshiv]

1.1-2.0 framerate? Is that good for this simulation?

[GyroCoder]

Very. How often do you have 27000 physical objects in one game at one time, all interacting with eachother at once?

[darkshiv]

All the time, with my super overclo-GAHHHHHHHHHHHHHHH! *Holds heart*

[c3h89]

A quad-core intel processor couldn't touch this. I believe this is one of the factors in the Futuremark 3d Mark pc test. What was your fps?

[docgeeves]

thats wat happened to the world trade center but taller

[roidroid]

yes, a simulation of a building collapsing would indeed be interesting.

i'm not sure why your comment was voted down, it's true - a simulation of the WTC collapse WOULD indeed be very useful.

[Everlong16]

You can sometimes set the fps so the objects are all combined in to a single fps, so it runs a lot faster and smoother than having a seperate fps for every object being rendered.

You can do it on the Crytek engine.

[kingnarutorule]

oooooh

balls :3

[00Kirby0]

Me want!

[feretz]

Ppl that whine at the fps, get garry's mod 10 and spawn 10k props and see how much fps you get...

[megladon06]

wow i have fps1 and mines is slower

[JerezJulio]

um?

[PelzaK1]

what cpu do You have?

[JerezJulio]

A Pentium e6600 2.4 ghz, However that demo runs the same in my old hyper threaded Pentium 2.4

This is because with that number of objects it is the system memory speed that determine the final performance.

To make scenes like that runs at in real time we need general purpose CPUs with lots of cores and also lots of shared local memory.

[mattlax8]

1 fps is horrible, no offense, i know no other engine can do this, but still 1 fps is horrible

[JerezJulio]

I know it is not as fast as I would like it to be.

If it makes you feel better I say the physics simulation alone is about 3 to 4 fps but rendering 27000 dense spheres one at a time is too slow in open gl, so the combined fps comes to be 1.1 fps.

Maybe in the near future when 8, 16 and 32 cores from Intel become available, scenes like this will run in real time, for now it can only run at iterative times in

[vmelkon]

Maybe pseudo instancing can be useful in this demo or if you can cull some of the objects with occlusion queries.

[bakura10]

Incredible video Julio ;).