News Raytracing in games not so far off

[Da Dego]

http://www.bit-tech.net/news/2006/08...ot_so_far_off/

[rupbert]

[mclean007]

Most raytracing actually traces the rays in the opposite direction, i.e. from the 'eye' back to the source, reflecting etc. all the way. The problem with this is it tends to leave things with quite hard shadows. The most impressive rendering technique I've ever seen is based on radiosity, which can produce beautiful soft shadows and some amazing subtle effects. For example, if you have a rubber ball on a matt white surface, the surface under the ball will take on a hint of red from light diffusely reflected off the ball. In contrast, in a raytracing engine, you might have some kind of soft shadowing, but most likely you'd have a red ball on a pure white surface with a hard shadow, and the whole thing wouldn't look too convincing.

Problem is, radiosity rendering takes a heck of a lot more power to process than ray tracing - on a modern PC it chugs along like the old raytracing experience on a 10 year old PC.

[rupbert]

@ mclean007

Interesting comments.

[DXR_13KE]

this is rather nice... more realism is good, but less playability is bad.

[Fod]

um, maclean007, radiosity is the method used in most realtime rendering engines. the main complexity in radiosity stems fromm computing the proportion of energy transfer between one patch on a surface and the other. for a static scene though, this never changes, so can be precomputed and stored in a data structure associated with each patch in the scene. this can be further optimised by using several other methods, at the cost of accuracy of course. even so, a fine-grained radiosity solution is not ideal, as you can;t do specularity without cheating, and often the shadows are not correct. they look nice, but they;re not accurate.

ray tarcing gives hard edges for shadows, yes, but only where the model has been simplified, either for a student's exercise or just to run faster. taking scattering into consideration increases the complexity of the solution by several orders of magnitude, and is essentially unfeasible. you can get around it by cheating a bit though.

[Tyinsar]

Good stuff again. For me this sheds new light on the AMD/ATI merger story: http://www.bit-tech.net/hardware/200...t_for_4/1.html

Could Intel be aiming at suddenly become THE graphics leader through a paradigm shift in graphics?

[SouthernBelle]

Wow. You sure are churning out the articles, eh?

[Agent_M]

intel is the leader in terms of sales :P

all this stuff is nice and all but id much rather have a game that uses my computer/consoles power for gameplay then has all the fancy stuff in the video sequences, like say final fantasy did before 10.

its nice that their pushing modern tech to its limits but really, i cant even play crysis on any good settings because my pc doesnt match the min specs and i dont have £600-700 to upgrade for a point and click game with pretty graphics.

the most fun ive had recently game wise was replaying fallout1 and 2 because of the great story and freedom that you have throughout the story. why cant they make intresting games anymore?

[Bindibadgi]

I was gonna say, this should make you well happy, Brett

[Cheap Mod Wannabe]

For Da Dego we just need to make a playroom game. Where bricks are actually bricks, and raytracing is used. Then he'll be able to happily spend time looking along the walls and not see textures being smeared. =)

[specofdust]

Now the core thing is exceptionally intresting. I've argued against the idea of excess cores being used for graphics in the past, but it sounds like raytracing would benefit hugely just from the extra CPU processing power a few more cores dedicated to it would bring. Might we see this coming into games big time during the DX10/Vista era?

[yahooadam]

interesting, but atm 7.5/30fps is a little low

[Faulk_Wulf]

Originally Posted by Agent_M

intel is the leader in terms of sales :P

all this stuff is nice and all but id much rather have a game that uses my computer/consoles power for gameplay then has all the fancy stuff in the video sequences, like say final fantasy did before 10.

its nice that their pushing modern tech to its limits but really, i cant even play crysis on any good settings because my pc doesnt match the min specs and i dont have £600-700 to upgrade for a point and click game with pretty graphics.

the most fun ive had recently game wise was replaying fallout1 and 2 because of the great story and freedom that you have throughout the story. why cant they make intresting games anymore?

Actually I believe that this would push INGAME graphics up to the level that most CG-Cutscenes are now. So that you'd essentially be playing a CG-Movie in terms of quality graphics.

Being on a budget myself, I do understand not being able to play games on a full capacity. I can't even play F.E.A.R. on full settings, and can't even run Doom III. But still, I think that this is still a cool step forward.

I won't dirrectly argue about games being fun. Personally I love FPS, but still boot the Command and Conquer games whenever I can.

This would be my 2c.

[LVMike]

Originally Posted by mclean007

Most raytracing actually traces the rays in the opposite direction, i.e. from the 'eye' back to the source, reflecting etc. all the way. The problem with this is it tends to leave things with quite hard shadows. The most impressive rendering technique I've ever seen is based on radiosity, which can produce beautiful soft shadows and some amazing subtle effects. For example, if you have a rubber ball on a matt white surface, the surface under the ball will take on a hint of red from light diffusely reflected off the ball. In contrast, in a raytracing engine, you might have some kind of soft shadowing, but most likely you'd have a red ball on a pure white surface with a hard shadow, and the whole thing wouldn't look too convincing.

Problem is, radiosity rendering takes a heck of a lot more power to process than ray tracing - on a modern PC it chugs along like the old raytracing experience on a 10 year old PC.

I use radiosity when im doing illuminations for my engineering class. If i take a room that is say 20X20 with a couple of sky lights, fully textured walls, and floors, paint ect, it takes about 10-20 hours on my A64 to render out the full sceen. We Use it like you said to get a realistic image of what the area will look like and the dispertion of light and shadow, into the corners and around objects, and the full spectrum radioscipy numbers, showing us wave lenght and amount of radiation.

Its way out there for on the fly stuff, but i cant wait for the day.

[Reaper_Unreal]

Haven't I been saying exactly this for the past year? Either way, having myself written a raytracer, I can atest that it isn't that far off graphics wise. Considering that there's a program that'll use current GPUs to raytrace 4d fractals at a decent resolution fairly quickly, I'd say were going to start seeing more raytracing happening in 2 years or so.

Though technically we're already using raytracing in certain games. Lots of visibility testing involves some sort of raytracing. Parallax Mapping is in fact raytracing, it's how terrains are done in raytracers (ish).

[specofdust]

But Radiosity is something that can be easily parrelised isn't it? So while it maybe takes a day on your A64 it could take merely 3-4 hours on one of the more highly clocked upcoming quad core Intel's. Since the natural progression is doubling(in computing that is), a few years down the line we might be at 8 or 16, by that time we'd be talking 1 or 2 hours, maybe considerably less if the CPU's improve in ways other then quantity. Things like radiosity and raytracing as a whole may be a little way off yet, but since they can aparently be parrelised "embarresingly easily"(according to John Carmack) then I can't imagine that we're more then a few years away from having the power to do these wonderfull things.

[Tyinsar]

Originally Posted by Agent_M

intel is the leader in terms of sales :P

uh, yeah, I think we all knew that. How's their graphics performance though?

What I was trying to say was if Intel can shift graphics back on to the CPU (something I was sceptical of before reading the article) they could become the leader in performance too. Maybe acidfire's idea http://forums.bit-tech.net/showpost....0&postcount=46 isn't too crazy after all.

[LVMike]

Originally Posted by specofdust

But Radiosity is something that can be easily parrelised isn't it? So while it maybe takes a day on your A64 it could take merely 3-4 hours on one of the more highly clocked upcoming quad core Intel's. Since the natural progression is doubling(in computing that is), a few years down the line we might be at 8 or 16, by that time we'd be talking 1 or 2 hours, maybe considerably less if the CPU's improve in ways other then quantity. Things like radiosity and raytracing as a whole may be a little way off yet, but since they can aparently be parrelised "embarresingly easily"(according to John Carmack) then I can't imagine that we're more then a few years away from having the power to do these wonderfull things.

Yes it does scale well to additional cores. Thats why the big boysin this stuff are using beowolf clusters to do large scale radiosity. It will come to mainstrem use. But like raytracing its going to take some time. But when it does... oh boy its going to be nice. Imagin imagery that is rendered in the same way that photons react to real world enviroments! It will create a level of realism that is only being glimpsed at today.

[EK-MDi]

Originally Posted by Bit-Tech

Clearly, we're not going to be seeing commercial games running raytrace for 2006, or likely 2007. But what about 2008?

I predict that we'll have raytracing early next year in 2007. In a little upcoming game called "Crysis", which will be one of the first games to use DirectX 10.