Nvidia Demonstrates Interactive Ray-tracing

[demonhorde665]

[citation][nom]matt87_50[/nom]very truethats why triangles are lame. they are for lame rasterizers. all shapes in raytracing should be formed from much more complex geometric objects for instance: terrain: should just a the equation for the plasma fractal you would otherwise use to generate a hight map. and bam! like 10s of bytes for your whole terrain! pwned!I mean, the ray collision equation might be a *bit* more complex.... but Tflops!! like how they are trying to overcome the same memory limitations in realtime rendering with tessellation.[/citation]

might i remind you that triangles are still the basic breakdown of ANY and ALL 3d rendering even when you add tessellation thea dded polysstilbreak down into triangles. rasterization has more to do with texturing than it does actual modeling so putting the two together really is inaccurate picture of things, that said ray tracying (atleast the reflections/refractions for surfaces also deals more with texturing than modeling. point is you sound silly and uninformed saying "triangles are lame" becaus4e even ray traced 3d uses triangles for the models ray tracing is about lighting and texturing not modeling , same for rasterization it's about texturing.

 

[dEAne]

I wait what ATI say for this. The images is so clear.

 

[matt87_50]

[citation][nom]demonhorde665[/nom]might i remind you that triangles are still the basic breakdown of ANY and ALL 3d rendering even when you add tessellation thea dded polysstilbreak down into triangles. rasterization has more to do with texturing than it does actual modeling so putting the two together really is inaccurate picture of things, that said ray tracying (atleast the reflections/refractions for surfaces also deals more with texturing than modeling. point is you sound silly and uninformed saying "triangles are lame" becaus4e even ray traced 3d uses triangles for the models ray tracing is about lighting and texturing not modeling , same for rasterization it's about texturing.[/citation]

yeah, your right, for standard video cards, triangles are all they do, even tessellation just breaks down into many triangles, but with raytracing, you can do what ever the hell you want. as long as you can collide with it, and get the tangent, your set. you can also map textures however you want.

for instance, at least for collision, a sphere is simpler than a triangle! certainly, it is simpler than the hundreds of triangles you would need to emulate a sphere, and even then it wouldn't be perfect.

it may be true that just processing triangles and optimizing for that is much faster than any other geometry, and yes, it certainly makes MAKING the geometry simpler, but if memory constraints are your major bottle neck (which it is), it might not be as crazy as you think to consider raytracing more complex geometry natively.

also, rasterizing is simply the process of taking mathematically defined geometry (vector graphics, triangles ect) and converting it to a grid of pixels, so ray tracing kinda falls under that, but generally it refers to the method that graphics cards use, where conceptually, it differs from ray tracing in that, ray tracing takes a pixel, and iterates all the geometry to find the color for that pixel, where as rasterizing takes a bit of geometry, then iterates all the pixels that need to be filled in based on that geometry.

 

[hardcore_gamer]

I'd like to see real-time ray-tracing in games..

 

[Chris_TC]

[citation][nom]lukeeu[/nom]Viper looks HORRIBLE for ray tracing!-No car shadow reflected in body-No road and shadow reflections on the rims-No reflection of the mirror in the windows[...][/citation]
The reflections are not selective. You see everything you're supposed to see from that angle.
They are using a ray depth of at least 4 because you can see the background through the side window and windshield.
Reflection ray depth may be less, but anything above 2 wouldn't be noticeable anyway in this type of scene.

And the scenery is obviously a photograph, probably mapped onto an environment sphere. You certainly cannot render it with "Oblivion level shaders."

 

[Chris_TC]

[citation][nom]demonhorde665[/nom]as far as i know of it ,no 3d apps make use of teh gpu past rendering teh basic view port , this is because when go to "render" a scene it appliies all the mapping technologies with teh cpu becuase a lot of those technologies can not be done on a video card (which uses rasterization) namely ray tracing , photometric lights, and ray traced shadows, by the time any of tehse tecnologies are doable by a video card ,they will likely have newer technologies that can't be rendered by a video card[/citation]
That couldn't be further from the truth. There are at least half a dozen GPU rendering engines available. iray, Arion, Octane, Vray RT and a bunch more. GPUs are so much faster than CPUs that this technology will grow more and more.

 

[godmodder]

Real-time raytracing: yes, real-time physically correct global illumination: NO! Such images take hours to render correctly, even today. And I don't foresee that to change anytime soon. Sorry Toms, but if you believe this is real-time global illumination, you've been misinformed.

 

[Chris_TC]

[citation][nom]godmodder[/nom]Real-time raytracing: yes, real-time physically correct global illumination: NO! Such images take hours to render correctly[/citation]
Why do you post when you clearly don't know what you're talking about? I use Octane Render from Refractive Software on a regular basis. It is an unbiased rendering engine using CUDA GPUs. Only one of many available on the market.

Unbiased means physically correct path tracing which includes full global illumination, reflection, refraction, caustics etc.

http://www.youtube.com/watch?v=xqJyD7M_WQg

 

[tester24]

is it me or does the pictures look grainy?

 

[lavos]

It's amazing that the human brain seems to operate at sub 200hz, yet it is possible to imagine seemingly photo-realistic graphics in realtime, and in the case of savants that detail can be brought into paintings in short amount of time.

 

[godmodder]

Why do you post when you clearly don't know what you're talking about? I use Octane Render from Refractive Software on a regular basis. It is an unbiased rendering engine using CUDA GPUs. Only one of many available on the market.

Unbiased means physically correct path tracing which includes full global illumination, reflection, refraction, caustics etc.

Why do you post such foolish remarks?
All these videos demonstrate only raytracing on models that wouldn't even need raytracing for realism in the first place! The model you showed me could be rendered with the same amount of realism with regular rasterization. Also, that "physically correct" renderer of yours only does 2-3 bounces of indirect lighting. Couple that with the relatively low resolution of this render and the coherent mix of materials and even I could code up a raytracer in CUDA that could rival this performance. Actually this video shows nothing new, it's just a mix of GPU raytracing and progressive refinement instead of waiting for the whole scene to be rendered correctly. There has been no breakthrough in global illumination rendering for years! Show me some footage of LOTR-style graphics rendered in real-time and then I'll believe you. Raytraced cars with so-called global illumination just make me angry.

 

[Chris_TC]

[citation][nom]godmodder[/nom]Also, that "physically correct" renderer of yours only does 2-3 bounces of indirect lighting.[/quote]
And once again you post even though you have no clue. The engine is obviously not limited to "2-3 bounces." The default setting is 8 bounces and you can crank it as high as you want. Not that it makes any visual difference after a certain point.

Show me some footage of LOTR-style graphics rendered in real-time and then I'll believe you. Raytraced cars with so-called global illumination just make me angry.[/citation]
Take a look at the gallery then if you want to see something other than cars.
http://www.refractivesoftware.com/gallery.html

What do "LOTR-style graphics in realtime" have to do with anything? CUDA rendering gives you a massive speed-up over CPU rendering. And that's the breakthrough. Nobody every said you could render everything in real-time.

 

[godmodder]

Did you look at the video of that interior rendered with global illumination? Although it was just a couple of indirect lighting bounces, it took ages for the image to become somewhat noise-free. To get a good quality image out of that would still take hours, so no breakthrough there. Let's break a common myth here: GPU raytracing isn't necessarily faster than a highly optimized CPU raytracer. In fact, the architecture of GPU's must be some of the worst ever to create a raytracer for. Like someone already said above here: it's not calculation power that matters here, but memory bandwidth and caches. That plus the fact that the GPU just cannot cope with unpredictive branches better than CPU's. Raytracing all but direct lighting requires shooting an exponential amount of incoherent rays that cause a HUGE amount of cache misses. Someone should solve the problem of making raytracing more coherently access memory data. Thèn we'll have a breakthrough. One at the algorithmic level and not just some constant increase in hardware power.

 

[randomizer]

iRay was supposed to be out in November last year and they're still just demonstrating it?

 

[Guest]

>And the scenery is obviously a photograph, probably mapped onto an environment sphere. You certainly cannot render it with "Oblivion level shaders."

It's not a photograph, its rendered scenery, that's the whole point of the demo!

 

[randomizer]

>And the scenery is obviously a photograph, probably mapped onto an environment sphere. You certainly cannot render it with "Oblivion level shaders."

It's not a photograph, its rendered scenery, that's the whole point of the demo!

Er... no, it's a photo. The car is rendered.