Raytraced games in 2 to 3 years, says Intel

[Amr-3DT]

PCGames Hardware:Are there game developers that already want to work with Raytracing?

Dr. Michael Vollmer: We keep in touch with companies all over the world - I dare say that in two to three years time we will see something. There already are some individual approaches, especially in the science sector, which show that Raytracing algorithms are scaling very well with the numbers of cores. But the migration to a new programming technology takes years; Raytracing is still in an early stage.

Read more: http://www.pcgameshardware.com/aid,6...rs_says_Intel/

[Kage]

Its a good technology, but even though it will allow for more advanced effects we see in films, etc, when it first starts out, graphics will be worse than what we have now.

[MadJester]

Originally Posted by Kage

Its a good technology, but even though it will allow for more advanced effects we see in films, etc, when it first starts out, graphics will be worse than what we have now.

He is more focusing on the real-time application of Ray-Tracing. Now a fully featured real-time previewer for the various packages out there would be nice. Maya and LW (though F-Prime) have some real-time Ray-Tracing but nothing to write home about.

On the article, Intel loves to pull this idea out and cart it around now and then but NVidia always ends up shredding their argument to bits a couple of days later. http://www.pcper.com/article.php?aid=530

Not to say that real-time ray-tracing won't happen but, for the foreseeable future, raster will continue to be used by games.

[jjp]

Although it certainly does have interesting applications.. for games I don't think anytime soon a change will happen. You can fake almost anything with rasterization (take a look at the latest movies from Pixar) and computationally it is a lot cheaper. Games usually don't consist of large sets of reflecting spheres.

My suspicion is that Intel does this for two reasons:

- Ray-tracing in games would be one of the very few cases in which a cpu with many cores would be useful in a consumer PC.
- They don't have decent rasterization hardware and probably can't compete with NVidia and ATI/AMD for some time.

[gnome]

Originally Posted by Kage

Its a good technology, but even though it will allow for more advanced effects we see in films, etc, when it first starts out, graphics will be worse than what we have now.

I don't think a (successful) game with raytracing will look significantly worse, because raytracing may sound cool to tech geeks, but I don't think your average gamer will give much of a damn whether it's rasterized or raytraced.

Originally Posted by MadJester

On the article, Intel loves to pull this idea out and cart it around now and then but NVidia always ends up shredding their argument to bits a couple of days later. http://www.pcper.com/article.php?aid=530

Not to say that real-time ray-tracing won't happen but, for the foreseeable future, raster will continue to be used by games.

hehe, nice statement by Kirk.
Rasterization with ray-casting pixel shaders or so sounds pretty neat, wonder how long it takes until we see something like that. (Or is there already something like it? )

[Kage]

I'm kind of confused as to wheather there is a really significant advantage over rasteration, even after reading that. It just seems like another technology that can produce the same work, with the same quality, but done realistically, without using any fake effects, like for instance real ambient occlusion instead of the fake stuff thats happening with games like Gears Of War 2, etc... But, that technology works, fake or not. So... is there really much point yet?

[MadJester]

The point would be in scalability. Say you have a dozen reflective orbs. Each time one of those orbs reflects in all the others you have to compute a cube map as in Portal with the many layers of portals. In Portal (if you played it) you might have noticed an option to limit the "depth" as in the number of times a portal would repeat the mirrored image. With raster every reflection is recomputed but with ray-tracing each portal or mirror is simply computing along each ray just like any other surface.

This idea of scalability comes into play in a number of places, reflective objects, distance calculations, shadow complexity and more but ray-tracing is way, way more system intensive overall. So ray-tracing would need to have significant gains in all the more specialized applications before it would become useful for an entire engine. As it stands now, looking at the article, the normal running and gunning around an environment runs at a max of 60 frames a second with ray-tracing and even that requires enormous processor power.

Rasterization with ray-casting pixel shaders or so sounds pretty neat, wonder how long it takes until we see something like that. (Or is there already something like it? )

Very true, I have no idea if there is something like that yet. If there is whoever has it is pretty mum about it. I can imagine a company making that the sole selling point for a game :-).