Intel: Some example of Ray Traced gaming likely in 2-3 years

by Ross Miller { Jul 31st 2008 at 9:15AM }

Already looking to the future, Intel's Dr. Michael Vollmer expressed hope towards the future of ray traced visuals in gaming. When asked by PC Gaming Hardware if there were developers already working on ray traced games, Vollmer said, "I dare say that in two to three years time we will see something," noting that he keeps in touch with companies all over the world.

Ray tracing is a graphics technique that is capable of some impressive, photorealistic visuals but conversely is very computational-intensive. Vollmer also said that the raytracing demo shown last month should be seen as a technical demo and that the graphics portion had been taken from the non-raytraced optimized Enemy Territory: Quake Wars. Speaking of which, we couldn't think of a more likely candidate for showcasing raytraced graphics than John Carmack and Id Software, except maybe Crytek. If Vollmer's estimates are accurate, we hope to see some ray traced frag fest in time for the 2011 Game Developers Conference.

[Via Blue's News]

Tags: crytek, id-software, intel, michael-vollmer, ray-tracing, raytracing

Reader Comments (Page 1 of 1)

Shadow Hog @ Jul 31st 2008 9:34AM

Wasn't the old Wolfenstein 3D engine based on raytracing? If so, this guy is 16-or-so years late.

Shattercross @ Jul 31st 2008 9:40AM

Wrong type of raytracing, this type of raytracing simulates refractions in glass and other clear objects.

WhoMe @ Jul 31st 2008 9:53AM

You are thinking of ray casting, which is not recursive up the wazzoo like ray tracing...

Tiptup300 @ Jul 31st 2008 11:27AM

Ray Casting: http://en.wikipedia.org/wiki/Ray_casting
Ray Tracing: http://en.wikipedia.org/wiki/Ray_tracing_(graphics)

Schnoog @ Jul 31st 2008 11:36AM

Wolf3D used something called Raycasting. Similar but MUCH faster. It placed constraints on the environment and only needed to fire a single ray per vertical column of pixels. Then using the distance to the intersection it would scale the size of the wall.

Habiib @ Jul 31st 2008 9:36AM

Someone visited Phatt Island.

SoulBlade @ Jul 31st 2008 9:37AM

Show me the next Myst!

JoeTheBlow @ Jul 31st 2008 9:39AM

Possibly, but incredibly unlikely.
Without massive amounts of power to spare, the incredibly complicated maths required are a huge price to pay for slightly better visuals (that a good tech team can "fake" using conventional techniques).

Its also been done on the PS3, without using the graphics chip (as Linux has no access to it):
http://eric_rollins.home.mindspring.com/ray/ray.html
http://www.youtube.com/watch?v=oLte5f34ya8

Nate @ Jul 31st 2008 9:47AM

Slightly better visuals?

http://en.wikipedia.org/wiki/Image:Glasses_800_edit.png

The visuals are a hundred times better. They just take forever to render.

I don't think we'll see any real-time games using raytracing for another decade. Once processors are 4-8 times as fast as they are now, and cheap enough to put 32 in a console (PS4 anyone?) then we might see this come to light, as it is a much more accurate method for generating photorealistic images, it's just also a couple orders of magnitude more processor intensive.

dextro @ Jul 31st 2008 9:50AM

But if I'm not mistaken Intel's great push with larabee will be to make feasible to use real-time ray-tracing. Just watch (and listen) to the video, you can make out some ideas on how to make real-time ray-tracing possible (like loosing a certain amount of pixels per reflection to keep the count workable for instance).

Will it take some time? Yes. Is it beyond reach? I don't think so and we may see it come true sooner than we might think (we will probably need different hardware than todays gpus though).

GenBanks @ Jul 31st 2008 9:51AM

The screenshot in the picture isn't 'slightly better' than what we're used to... Unless you mean slightly better than what rastarization (or whatever it's called) will reach in the time it takes for ray-tracing to be feasible with the hardware that will be out.

PunchyMcJunk @ Jul 31st 2008 12:09PM

MATHS:

Mathematical
Anti
Telharsic
Harfatum
Septomin

http://www.youtube.com/watch?v=0F3oxzAcNNE

BananaBoat @ Jul 31st 2008 12:27PM

Wow...apparently it's possible to notice the uncanny valley in glassware and dice. Then again, if that were being rendered in real time on a table I wasn't focusing on, it would look ridiculously real. When the GTX 260 I just bought though (got it on sale for 200, wouldn't have payed regular price) won't get me a steady 30 FPS in Crysis with any samplings of AA enabled (without it runs like a dream, and it isn't really needed, but still) at 1080p, I don't think the graphics card market will get anywhere close to Ray Tracing any time soon.

Hardcore @ Jul 31st 2008 9:45AM

Indeed, with huge computational power needed for raytracing, I doubt will this feasible in near future, unless there are very significant changes made in hardware.

Haggard @ Jul 31st 2008 9:48AM

I imagine this might make up for the fact that PC hardware is evolving far faster than graphics in games are.

But give it a few years from its debut and Ray Tracing is gonna kick major ass.

Zeromaru @ Jul 31st 2008 9:50AM

Another, far more surprising aspect to ray tracing is that it's mathematically much simpler and easier to program.

http://www.superjer.com/pixelmachine/

This is a weekend project that one guy undertook to write a ray tracing renderer, end result is just over 800 lines.

The reason why it's so photorealistic is because it works just like a camera. It determines what color each pixel should be by firing a photon and seeing how it reflects (basically calculates the reversed path a photon takes to get to your eyes).

WhoMe @ Jul 31st 2008 9:59AM

"Another, far more surprising aspect to ray tracing is that it's mathematically much simpler and easier to program."

This is true, to a certain extent.

You can write a simple and functional ray tracing algorithm that fits on a business card. However, I wouldn't say they are all "much simpler and easier to program". Once you know the actual algorithm, it is easy to implement, but that is the case for virtually any piece of software.

The big thing in global illumination or ray tracing is coming up with an _efficient_ algoritm, some of which use some _really_ fancy math.

WhoMe @ Jul 31st 2008 10:04AM

Also, those images are not photorealistic. Not in the slightest. Looks to me he's using Whitted style ray tracing. Which is indeed very easy. But it does not deliver photorealistic imagery at all (think hard shadows and no soft phenomena).

What it does do well is fancy chrome spheres, incidentally something that is always present in a demo scene for stuff like this. :p

Shaun @ Jul 31st 2008 9:51AM

That would be a massive massive advancement in terms of graphics.

WhoMe @ Jul 31st 2008 9:53AM

I did my thesis on global illumination. Neat stuff, but the problems are still ginormous. I'm not saying it will never happen, but I believe it's gonna take a good long while before any major shift happens.

A lot of these technologies are still so new. The algorithms aren't really up to snuff yet for real time stuff. Especially considering games (animations) breaks a lot of optimizations. Hardware isn't there yet either. Although you can exploit current programmable hardware. A lot more research needs to be done and the hardware needs to be there.

What we have now is simply "good enough" I'm afraid, not to mention a lot of money and effort went into the current technology. Both algorithms and hardware need to evolve to make real time global illumination feasible.

"Speaking of which, we couldn't think of a more likely candidate for showcasing raytraced graphics than John Carmack and Id Software, except maybe Crytek."

Why would these companies be a likely candidate? These companies are heavily invested in traditional rendering. Sure, Carmack is a graphics genius, but I think he'll take a more practical approach to this, rather than do it just for the sake of it. I would sooner expect something to come out of left field, and then when this matures, a slow shift.

A @ Jul 31st 2008 3:52PM

I think theyre good candidates- Carmack has always been very devoted to the newest techs, and very excitable where theyre concerned. If he gets interested in ray tracing, I can't wait to play Doom4, monster closets or not!

Also, Crytek just makes gorgeous stuff. so, yeah. that's all i got.

Vladeon @ Jul 31st 2008 10:05AM

Oh, so that's what an updated, "Day of the Tentacle" game would look like... Though, I don't remember the part where you get drunk in a bar, must've been when you were talking to George Washington and Thomas Jefferson.

fred @ Jul 31st 2008 10:21AM

everyone here knows so much more about maths and tracing than me

:[

I don't understand all your words

minus_273 @ Jul 31st 2008 10:23AM

but but, killzone 2 is ray traced! its the power of bluray and cell! the devs claim it uses ray tracing:

http://www.messageboardchampion.com/?p=15

the power of cell compels you

knighty @ Jul 31st 2008 10:23AM

The problem is that to get any decent performance we'll have to think about totally different graphics hardware architecture, since at the moment ray tracing has to be done purely by CPU which is phenomenally slow.

BeeKnuckle @ Jul 31st 2008 10:31AM

Back in 1994 Nintendo Power was all about some ray-traced graphics in Rise of the Robots. That was all they could say about the game too. It sucked.

edgore @ Jul 31st 2008 11:30AM

That is because they were using ray tracing to create static bitmap sprites and then throwing those up on the screen, whih is generally going to look pretty awful, especially once they have been downsampled to the relatively low resolution of a 16-bit console. The concept here is to actaully render the images in realtime which, if possible, would look much, much, much better.

samfish @ Jul 31st 2008 12:46PM

Rise of the Robots 2 in 2019 CONFIRMED!

j.howlett @ Jul 31st 2008 10:36AM

nice

Pidge @ Jul 31st 2008 11:31AM

This technology shows a lot of potential.

Brent @ Jul 31st 2008 11:50AM

I wrote a ray tracer once. It took a few hours to render a scene, especially if I did anything fancy like depth of field, multiple light sources, etc. I would say it is probably not worth it...current gen games look pretty damn good to me.

nick @ Jul 31st 2008 12:12PM

The picture above is from 2006

dark_inchworm @ Jul 31st 2008 12:15PM

This post contains a decent amount of comments that required Genuine Thought. What the fuck, people? :)

Dr. Stabbingworth @ Jul 31st 2008 1:03PM

It's because people with anything legit to say are drowned out by idiot fanboys who have nothing better to do but wait for every update and spam the site with garbage.

I still remember when epobirs and laughing target would write epic posts. This place was great.

At least Vlad is gone.

J @ Jul 31st 2008 1:11PM

No doubt... Here's to hoping that just posting here will boost my IQ a, much needed, point or 2.

Aaron @ Jul 31st 2008 12:28PM

http://www.youtube.com/watch?v=oLte5f34ya8

Real time interactive raytracing using a cluster of 3 PS3's and Linux.

Shawn @ Jul 31st 2008 1:02PM

just think how many PS3's they will need for a fully fledged game running at 1080P, with everything going crazy. That was 3 PS3's just to show light on a single car with nothing going on in the background. I wonder what the new GPU's from Nvidia/ATI plus Intel's new chips would be able to accomplish now.

WhoMe @ Jul 31st 2008 1:28PM

Yeah they are always quick to use rigid shiny objects in demos. While this car is probably a lot more physically plausable than a car from GT for instance, GT has about a dozen or more of them on the screen simultaneously which look damn good as well.

Aaron @ Jul 31st 2008 1:28PM

I think what all this means is that the future of gaming is going to rely very heavily on multicore architectures like the Cell. All these developers whining about how the PS3 is "too hard to program for" better start getting their act together, because it won't get any easier down the road. Graphics chips are going to get more advanced and have multiple cores. CPU architectures like the Cell will also be necessary for advanced graphics. The Xbots can poo poo the Cell all they want, but the reality is that the Cell (or something like it) is the future.

Shawn @ Jul 31st 2008 12:58PM

Wow, that pic is amazing, finally something that makes Crysis look dated :)

I wonder, since it's heavily rumored that Intel is getting into the high end GPU business, this could be what they're working on more then we think. I don't know much about Raytracing, but the biggest problem to me seems like it's going to be a power consumption issue. That's probably a major problem with all of this, trying to figure out how to achieve this with what is considered a "practical" amount of energy.

For programmers, is Raytracing more convenient or does it take more man-hours to achieve the same effect we have today? Bottom line, is it worth it over the current path we are on right now in heavy game development?

WhoMe @ Jul 31st 2008 1:24PM

In general, it is easier on the artist. GI algorithms generally closely approximate the real physics of light transport. So in theory, simply defining the correct material parameters for something and it's textures should yield very convincing results.

It's also a lot easier to achieve certain effects. Shadows, something which is notoriously difficult to simulate using traditional hardware, essentially come "free" with ray tracing. Same thing for other effects such as reflection and refraction.

Of course, you don't always want hyper realism (think special effects which provide cues for the player).

Everybody always mentions reflections and such as advantages of ray tracing for example, but let's not forget that lighting in current games in generally piss poor compared to the real deal. Although various techniques are being used to simulate indirect illumination, which has a sometimes subtle, but nonetheless huge impact on the perceived realism.

Lekko @ Jul 31st 2008 1:54PM

I have a feeling raytracing is what Sony was originally trying for with the PS3, but then couldn't quite make it work. If you look at the Cell architecture, it is designed for massive floating point performance, which isn't that well suited for a conventional CPU, or really for game code per se. You can make it work (Obviously they have), but that's not what it was built for.

When I look at Cell, I see a type of processor that raytracing will eventually run on. Cell acts much more like a GPGPU than a CPU. Kinda what Larabee is going to be. I think when Sony was working on the PS3, they wanted to just go for full realtime raytracing, found that the cell could not do it, and that the system would require far more power than it had. So then they were sitting on this chip they spent tons of R&D on, but didn't really do what it was supposed to. Enter Nvidia.

Raytracing, to me, looks closer than you would think. Since it is massively scalable, and incredibly parallel, processing is heading in that direction. With multiple cores becoming the norm, floating point performance needs only a few more short leaps before we have a system that could push it. Cell2 is supposed (rumored via roadmap) to be 32 SPU, with varied optimizations. I'm not one to bust out a calculator to see if that is enough to use cell purely as a GPU to render raytracing, but it could come close.

mjs494 @ Jul 31st 2008 2:42PM

"Carmack seems to think that Intel's direction using traditional ray tracing methods is not going to work and instead theorizes that using ray casting to traverse a new data structure he is developing is the best course of action. The 'sparse voxel octree' that Carmack discusses would allow for 'unique geometry down to the equivalent of the texel across everything.' "