750,000 particles in XNA

[fraserofthenight]

Wow the NVIDIA fanboy is strong here.

[NickEnchev]

love this song!

[carlitrosss]

This rocks.

Did u code this in c#?

[GearGOD]

Just a few things:

This video was posted 2 years ago and even at the time was running on very dated hardware. Today similar simulations can be pushed to an order of magnitude more particles.

I complained about ATI because their then-current generation of video cards didn't support VTF despite it being a compulsory part of the dx9 spec. All cards since then support it as far as I know.

I can't give you the source because I don't have it. I lost most of my XNA work with a broken hard drive.

[mazdaplz]

@GearGOD

no need for the code what was the general idea did you make a wind factor?

[david0aloha]

This is beautiful! You said the rendering is done on a Nvidia card. Is there any particular part of the hardware acceleration that is responsible for this. Is there a particular library you're using? I love the flow patterns - you get very real boundary layers that just don't show up in nearly any fluid/particle systems in real-time. There's got to be some sort of efficient wave-equation being used (from the looks of the discrete particle "waves").

I would love to play around with that.

[GearGOD]

@david0aloha

There's no wave equation at all here, its some very simple linear motion that was tweaked to look nice. Today you can do this on any reasonably modern graphics card, ATI or Nvidia. This was done in XNA with vertex texture fetch, but from what I understand DX10 offers an even better way to do this - check out the video response linked here.

[BlueRajasmyk]

@GearGOD What is VTF, and how in the world did you code this in XNA (HLSL?)

[msqrt]

Wtf is this ATi hating? It's not like only nVidias can run loads of particles.

[Arishava]

lol, give me this program!!!^^

[skate2spellgirl]

what FPS do you get for this, i see it a little bit sketchy

[ndrunkenTeddy]

nice

[QRebound]

Might I see the source code? I can credit you in my project. \o/

[DerangedGames]

Looks amazing. and looks alot like water!

[HotNoob]

Wow, that's so awesome...

Once im done making this 2D game with XNA ima head on to 3D and cool effect thingies :D

[CSMHD]

wow good work, nice gpu calculation.

[ilhadosmacacos]

thats really fucking cool!

pobabilly you have a nasa computer with physx 9000,1!

[guert2000]

I want that on my screensaver :)

[byteXI]

Beautiful..

[JanitorMaster]

Do you run this on a normal desktop machine?

[ninjafrank1337]

Wow, that's real time.

[mouser58907]

gorgeous! Any chance of seeing source code?

[digitaldud]

Yeah I just found out my shitty ATI-powered laptop doesn't have vertex texture fetch either. I should have gotten nvidia.

[BlooD2oo1]

I made a similar app too,

try it out in real time:

blood2oo1,atw,hu

[PontusM4gnusson]

Enter here to get virus!

[BlooD2oo1]

It's very nice!

[suggie1337]

Woa nice!

[bcapecci]

:o Wow nice! And thank you so very much ATI because this wouldn't be amazing or anything in commercial software.

[DamianProgrammer]

im guessing you did something like this:

(1 / distanceFromMouse) * gravitationalPull

[Dgamer5000]

Thats overly simplistic, but i do assume to something "like" that can be found somewhere in the code. To achieve something like this would need extensive use of vectors and vector operations to simulate the forces caused by the downwards gravity and the inward pull of the mouse cursor.

I don't think the math is all that impressive or would be that difficult to understand, but the polish is what makes this video complete utter awesome.

[DamianProgrammer]

The thing that puzzles me is how they fit 750k particles without lag, I made something like that, exept its only about 17k particles.

[Riebart]

Just looking at what he's got, my guess is that his "cursor attraction" force isn't like gravity (An inverse-square relation), but simply an inverse relation (Some things just don't happen in an inverse square relation).

I would be curious the FPS he is getting.

Exceptionally pretty though. :)

[SumN3rd]

Wow, That is amazing *watches again*

[Dgamer5000]

I want to have sex with this video.

[rogacz86]

Screw ses with it. I want to have babies ;p

[icoo44]

thats sick good job

[moitoius]

Textures in Vertex Shaders? Yeah, I know what you mean! My landscape is sooo much more complicated. I read a interview with them and they said "It would never be used." Reason why I recommend NVidia to everyone.

[arschochse]

Amazing!

btw whats the name of the song?

[TheNight999]

crystal method - born too slow

[DamianProgrammer]

im still wondering how you did that

[DamianProgrammer]

ooooooh, i see how you did it, the closer the particles are to the mouse the quicker they acelerate toward it, i think, please reply

[SoraCross]

Oh, nice. I was wondering too. I'd +1 you but my firefox is broken on this PC.

[DamianProgrammer]

thx, lol?

[luketheduke0]

How did you make it efficient? I have tried stuff with particles, but I can't get them to interact efficiently.

[cyborgtroy]

I think with this many particles you'd want to use GPU shaders instead of using the CPU to calculate particle physics.

tl:dr version: I HAVE NO IDEA

[luketheduke0]

So you think it would be possible to use that GPU shader in something like XNA?

[luketheduke0]

Oh, dur-da-dur, it says that on the title.

[GearGOD]

Of course its possible.

A floating point texture is used to represent particle data. Each pixel represents a position and velocity. The texture is drawn to itself with a shader that updates the acceleration/velocities.

Meanwhile there's a prefab mesh with a bunch of quads is sent to the gpu with a vertex shader that does vertex texture fetch from the floating point texture. Voila, easy 2d system. A 3d one is a bit more fiddly as it requires more textures.

[xfxeduardointel10]

MAN!!! this is so AMAZING!!! very cool!.

what is your experience in c #? how long you program in C #?

very cool video man! good work!

[ezza297]

Dude this is truly amazing good work!

[kotsoft]

just wondering. what is this loophole in sm2? i have a gpu fire simulation working well on nvidia cards but not ati cards so i've been looking everywhere to find a solution.

[GearGOD]

Vertex texture fetch.

If you post your shader code somewhere I might be able to tell you if there's anything wrong-looking.

[kingcol252]

:30 was the best. where the hell can i get this?

[kraelen]

What system specs was this run on?

[GearGOD]

Pentium4 2ghz, gf6600. In other words, very poor compared to a modern computer. I've run the sim with over 1 million particles at 60+ fps on a newer computer.

[alexjo013]

holy shit! this was amazing!!!!!!!!!!!!! :O

[dexter2k6]

One of the coolest effects i´ve seen so far, codet with xna!

[stuart12926]

wow great particles !!!!

[Enigma415]

Holy fuck that's sweet!

Impressive algorithm, too!

[sorrow1989]

im not a big fan of ATI, but this is very impressive

[bloodelfpallie]

FUCK ATI (why wont they let it be used)

[GearGOD]

ATI used a loophole in the shader model 2 spec to avoid having to support this even though its mandatory for any card claiming to support sm2.

[bloodelfpallie]

arrgh someone should make a petition

[AndrewHeaford]

Wow, that's very shiny.

Very.