News Nvidia: Ray tracing is too intensive for Hollywood

http://www.bit-tech.net/news/2008/08/26/nvidia-ray-tracing-is-too-intensive-for-hollywood/1

NVISION 2008: The ray tracing and rasterisation debate came up during a chat I had earlier... Nvidia says that ray tracing is too much for Hollywood.

 

Have they got any new products in the pipeline or are they just going to tell us how great GT200 is for three days? I'm wondering because there doesn't seem to be an interesting schedule for this thing ala IDF..

 

So... Ray tracing doesn't look better?

 

There's a few things bugging me...

1. Pixar has developed it's software in house for a long time, it's called renderman and although Nvidia can claim some fame from it, Pixar's programmers get the real pat on the back. It's probably the most intense program there is, next to a few othere renderers.

2. Raytracing has been used for a long time (in the computer sense) in Hollywood and is the most realistic way of doing anything, lordy even I can do it!

3. As far as switching to whole raytrace realtime renderers for videogaming, that's going to be a slow proccess.

 

Ray Tracing has been used in Hollywood, but only for a small part. A lot of things are still done in rasterisation. ( Even though how many article tell you something is Ray traced, a lot of other graphics are done in Rasteriation ) And it will take at least another 10 years before we even switch to fully Ray Traced World.

So why are Nvidia so sure of it? Because back in the days of Riva TNT and Geforce, they famously said they could do real time render of Ray Traced Toy Story in 5 years time. There were also PS2 at the time stating how powerful its emotional engine. But reality comes, 10 years later we are still NO WHERE CLOSE to ray trace gaming!

As we keep bumping up Resolution, more effects, Default to High Quality AA, rasterisation will still have a long way to go.

 

Shouldn't Resolution and AA benefit from Ray Tracing?...
Once the "ray is traced", so to speak, shouldn't matter in what resolution you display it, right. You calculate them down from the mathematic exact position anyway?

???

Xir

 

I don't get this. My understanding is that modern graphics engines are hybrid rasterisation/ray-tracing engines anyway - they typically draw a frame by taking the initial steps from a classical rasterisation algorithm (matrix transformation of vertices from world space to eye space; back-face culling; z-culling etc.), then render using pixel shaders, which will do lots of things, including environmental texturing and lighting using rays. I would expect that the underlying geometry will remain with the standard rasterisation route (i.e., fundamentally, translating triangles in world space into eye space) for some time, with more and more of the actual rendering being shunted over towards a more ray-tracing-esque system as hardware gets more powerful and pixel shaders get ever more complex in the hunt for more and more realism.

Is Nvidia saying that 'pure' ray-tracing (i.e. describing each object in a scene as a mathematical solid rather than as a collection of flat triangles, then rendering the whole scene using light rays and eye rays) is too intensive? Is it to do wtih global illumination? I have seen working realtime ray-tracing (e.g. Q4RT) and it was pretty impressive. Not perfect, mind, but not at all bad as a demo of what can be done in real time using only a CPU with no specialist hardware.

@Xir - one big plus of ray-tracing is that it scales amazingly well - increasing resolution is approximately linear with performance, as is adding processing cores. Pixel perfect shadows and reflections come 'free', and you can achieve excellent AA by oversampling edges.

 

Here it is for y'all
http://www.maxwellrender.com/
this thing is amazing

 

Nice link John, thanks