New player in the GPU market!

[Toss3]

SAN FRANCISCO - March 10, 2009 - Caustic Graphics®, a new 3D computer graphics company, launches today with a fundamental breakthrough in raytracing acceleration that is set to define a new era in professional 3D production and interactive consumer graphics. Raytracing, the gold-standard for creating 3D imagery, duplicates the natural physics of light, creating stunning images by meticulously tracing the path of light to and throughout any given scene.

Caustic's first-generation technology will deliver an average 20X increase in the speed used to create stunning, realistic 3D imagery for film and video, game development, as well as automotive and consumer product design. The second generation of Caustic's technology, due early next year, is expected to gain an additional order of magnitude in performance, offering 200X speed over today's state-of-the-art graphics products. This massive speed jump is due to Caustic's patent-pending raytracing algorithms implemented in a semiconductor design.

The computational complexity of producing cinema-quality, raytraced 3D images involves large, downstream costs, including slow "black box" design iterations and costly "render farm" server infrastructures. These costs are symptoms of a problem with today's computer designs where CPUs and GPUs are efficient at accelerating the rasterized graphics in video games but woefully inefficient at accelerating cinema-quality raytraced graphics. Caustic's forthcoming standards-based CausticRT™ platform enables highly parallel CPUs and GPUs to massively-accelerate raytracing, putting it on par with rasterization and resulting in cinema-quality 3D delivered interactively on low-cost PCs.

"Real-time raytracing has been the holy grail of computer graphics since 1979 - a dream always on the horizon but never within reach," said Dr. Jon Peddie, of Jon Peddie Research, the computer graphics market research firm in Tiburon, CA. "Demos have been done with 16 or more processors, super computers, and other esoteric devices, but never anything that was within reach of a PC budget. Caustic Graphics has made the breakthrough with a combination of a small hardware accelerator and some very innovative software to be able to deliver real-time, complex, high-resolution raytraced images - this is an amazing accomplishment." The Caustic management team is made up of technical visionaries and graphics experts from Autodesk, Apple, ATI, Intel and NVIDIA. Before starting Caustic, company founders James McCombe, Luke Peterson and Ryan Salsbury worked together at Apple, where McCombe was a lead architect for the company's OpenGL Graphics system and Chief Architect of Apple's rendering algorithms for the iPhone and iPod.

"For years, 3D professionals in multiple industries have labored under the yoke of slow iterations and unwieldy offline render farms," said Caustic Graphics CEO, Ken Daniels. "Caustic puts the power of a render farm, operating at interactive speeds, on every desktop, enabling designers and animators to get from concept to product faster, better and at lower cost." The Caustic product offering will be announced in April 2009.

About Caustic Graphics
Caustic Graphics, creators of CausticRT, is reinventing raytracing and changing how interactive cinema-quality 3D graphics are produced, used, and enjoyed. The company, headquartered in San Francisco, is currently funded by angel investors. For more information, please visit www.caustic.com.
http://www.caustic.com/company_press-releases.php

http://www.caustic.com/images/platform_chart.gif
http://www.caustic.com/dev_intro.php
http://www.caustic.com/images/pic_board_large.gif
http://www.caustic.com/caustic-rt_caustic-one.php


So real-time raytracing might just become a reality much earlier than expected and intel's larrabee is in for some stiff competition!

Too bad they're not yet aiming for the general consumer market. ;)

[CaptNKILL]

That's used for rendering 3D graphics for movies. Not real-time games.

[Galvin]

Good cut into nvidia's pro market share unless they come out with something that beats it.

[walterman]

Very interesting. We'll see fully raytraced games not so far.

[Toss3]

That's used for rendering 3D graphics for movies. Not real-time games.

Well yeah, but it's a step in the right direction and there's nothing that says this couldn't be used in games as well. :) Nvidia and ati are going to want to be able to compete with this, which means that the technology is eventually going to reach the consumer graphics cards as well.

[Bman212121]

Judging by the passive cooling on the processors it probably isn't all that powerful... It might be good for one purpose like a PPU, but it most likely falls way behind on anything else.

[CaptNKILL]

Well yeah, but it's a step in the right direction and there's nothing that says this couldn't be used in games as well. :) Nvidia and ati are going to want to be able to compete with this, which means that the technology is eventually going to reach the consumer graphics cards as well.

Maybe.

But I think its a bit premature to call them a new player in the GPU market. They aren't in the GPU market yet. They just make a ray tracing accelerator.

It sounds like it actually works along side other GPUs as well. It doesn't have any video outputs either.

This might have some effect on gamers in 5-10 years but I don't think ray tracing is a big priority right now.

[hell_of_doom227]

Another thing Nvidia should buy and implement in their GPU :)

[Toss3]

Maybe.

But I think its a bit premature to call them a new player in the GPU market. They aren't in the GPU market yet. They just make a ray tracing accelerator.

It sounds like it actually works along side other GPUs as well. It doesn't have any video outputs either.

This might have some effect on gamers in 5-10 years but I don't think ray tracing is a big priority right now.

It's probably just a render accelerator as you said(which would explain why it doesn't have video outputs), but they could make it into a desktop add-in card like ageia's physx solution.

Real-time ray tracing for games isn't that far off if you ask me, as games aren't as complex as most professionally rendered-scenes and as such wouldn't require as much rendering power. So if you could render a complex scene at 1 fps with the help of this solution, then games could probably be rendered at a much higher speed(playable framerates at least).

It all depends on whether directx 11 is going to feature ray tracing or not though because if the next generation of consoles won't have support for it then we'll have to wait another 5 years for it to become a reality(I know it isn't included, but they just might make it for dx11.1).

[CaptNKILL]

I don't think the problem is speed alone, its that everything that is used in 3D games now is done without ray tracing.

IMO, if you can get all of the lighting effects in games like Crysis, Far Cry 2, STALKER, Doom 3 or Chronicles of Rid**** at high frame rates without using ray tracing, there isn't really any need to completely reinvent the way that real time 3D graphics are displayed, just to go back to 1024x768 at 30fps.

It just seems unnecessary at this point. If they tried to push ray tracing any time soon it'd almost certainly fail to catch on.

This is a pretty good read:
http://www.pcper.com/article.php?aid=506&type=expert&pid=7

[Toss3]

I don't think the problem is speed alone, its that everything that is used in 3D games now is done without ray tracing.

IMO, if you can get all of the lighting effects in games like Crysis, Far Cry 2, STALKER, Doom 3 or Chronicles of Rid**** at high frame rates without using ray tracing, there isn't really any need to completely reinvent the way that real time 3D graphics are displayed, just to go back to 1024x768 at 30fps.

It just seems unnecessary at this point. If they tried to push ray tracing any time soon it'd almost certainly fail to catch on.

This is a pretty good read:
http://www.pcper.com/article.php?aid=506&type=expert&pid=7

Thanks for the link, but the most important thing here is that a full ray-traced game doesn't need a gpu, whereas rasterization needs a powerful gpu to display the same effects.

Remember the ray traced Quake 4 video? That was running at a 1024x1024 resolution at 100fps with 8 cores(the gpu only output the video). So if this code was optimized for say cuda - then just think about how amazingly fast a gpu could render scenes(there are already a couple of hardware ray tracers out there, but they are slow(this is where Caustic comes in!)?

Here (http://www.pcper.com/article.php?aid=530)'s a nice interview with Nvidia's David Kirk on rasterization vs ray tracing

[CaptNKILL]

Thanks for the link, but the most important thing here is that a full ray-traced game doesn't need a gpu, whereas rasterization needs a powerful gpu to display the same effects.

Remember the ray traced Quake 4 video? That was running at a 1024x1024 resolution at 100fps with 8 cores(the gpu only output the video). So if this code was optimized for say cuda - then just think about how amazingly fast a gpu could render scenes(there are already a couple of hardware ray tracers out there, but they are slow(this is where Caustic comes in!)?

Here (http://www.pcper.com/article.php?aid=530)'s a nice interview with Nvidia's David Kirk on rasterization vs ray tracing

I guess the thing that's stopping me from jumping on ray-tracing is that every single real time ray tracing demo I've seen looked horrible. :o

They add razor sharp projected shadows and perfect reflections but they remove everything that makes games actually look good. I'm guessing this is because current effects (shaders, normal mapping, soft shadows etc.) do not mesh well with ray tracing.

I just don't think the demand for ray tracing is anywhere near big enough to turn real-time graphics in a completely different direction.

Look at these demos for example...

DsCgJhoAm0c

2qzpl3XE6zs

mtHDSG2wNho

Compare these to the rasterized versions...

3rKKNnZJfjE

Tx5r2Jhp3XU

The ray traced ones pretty much look horrible. :o

If anyone is going to try to sell this technology to the public they need to give us some reasons why we should want it. Throwing mirrors in odd places isn't going to do it I'm afraid. I'm all for it if they can make the games look AND run better than they do now, but its looking like they're struggling to meet the graphics quality of 1999 just for the sake of using new technology.

[Atomizer]

I agree, all the raytracing demos have pretty much no graphical features that have been pushed out from DX8 onwards(shaders), the only thing it seems good for is creating extreme mirrored surfaces, i have yet to see them do REALISTIC reflections with raytracing.
I have heard arguments saying it can all be done in raytracing, but I submit to anyone who supports realtime raytracing: prove it.

[harl]

Any specs? (Bandwith, flops, ...)

[walterman]

The ray traced ones pretty much look horrible. :o

I think that the light in those games was not designed for ray tracing.

[CaptNKILL]

I think that the light in those games was not designed for ray tracing.

Until someone can show a game that is designed for real time ray tracing that looks better than standard rasterization, this stuff is just a pipe dream.

Real-time Toy Story with ray tracing is at least a decade away. Where as, you could probably get reasonably close with CryEngine 2 with decent performance on current hardware using pure rasterization.

[crainger]

Rast rast rast!

[jolle]

The carmack agrees:

CARMACK: I wouldn’t say there’s anything that Intel has shown, even if they network a whole room full of PCs and say “we'll be able to stick all of this on a graphics card for you in the coming generation,” I don’t think they’ve shown the win. I don’t think they’ve shown something people will say “my god that’s 10x cooler” or “that makes me want to buy a new console”.

It is tough in a research environment to do that because so much of the content battle now is media rather than algorithms. They’ve certainly been hacking on the Quake code bases to at least give them something that is not an ivory tower toy, but they’re working with something that is previous generation technology and trying to make it look like something that is going to a next-gen technology. You really can’t stretch media over two generational gaps like that, so they’re stuck. Which is why I’m hoping to be able to do my part and provide some proof of concept demo technology this year.

but there is very little in the way of real proof-of-concept work going on. There’s no one showing the demo of like, here this is what games are going to look like on the next generation when we have 10x more processing power - nothing compelling has actually been demonstrated and everyone is busy making these multi-billion dollar decisions about what things are going to be like 5 years from now in the gaming world. I have a direction in mind with this but until everybody can actually make movies of what this is going to be like at subscale speeds, it’s distressing to me that there is so much effort going on without anybody showing exactly what the prize is that all of this is going to give us.

From an older interview largely around raytracing.
http://www.pcper.com/article.php?aid=532&type=overview

They arent really showing anything that justifies the move to raytracing, why we must expend these huge amounts of power to enable this.
Obviously we have offline renders to look at as a goal, but as mentioned, the performance to do that is a long way away, softening raytraced shadows is expensive as hell for example (judging from the performance impact in Mental Ray atleast)

On that topic, he has some ideas of using raytracing for things other then rendering, mentioned in the interview.
Anyhow, thats ONE developers views on the whole raytracing thing.

[MaxFX]

nVidia bought some raytracing company in 2007 I belive and rest assured they are at heavy work to make their kind of hardware/implementation. this wont fly at all I think as nvidia and Carmac and other devs stated over and over.

It will all be some kind of hybrid rendering as going only raytracing is no good at all. it have to be rasterization their to as only playing with rays wont cut it as they said :D

[Dreamweavernoob]

Personally even though im going to be bashed for saying it :D, my preferred direction would be in the experience of computers. No matter the graphics at the moment, you are left interfacing with a keyboard and mouse via screen with no sensory enjoyment (except visual). Start incoporating feelings and experiences in technology.


*I just want a sex simulator ROFL. :o :D

[Atomizer]

Personally even though im going to be bashed for saying it :D, my preferred direction would be in the experience of computers. No matter the graphics at the moment, you are left interfacing with a keyboard and mouse via screen with no sensory enjoyment (except visual). Start incoporating feelings and experiences in technology.


*I just want a sex simulator ROFL. :o :D

I pretty much agree(Though I get a boner for awesome graphics >.> )
I am still waiting for decent mind powered devices like the various devices that have already been shown at various tech events, it will bring gaming to a whole other level, and since they are PC peripherals, will quickly get past the "gimick" stage, unlike another white console toy I wont mention....oh, too late <.<

[jlippo]

The carmack agrees:

From an older interview largely around raytracing.
http://www.pcper.com/article.php?aid=532&type=overview

They arent really showing anything that justifies the move to raytracing, why we must expend these huge amounts of power to enable this.

There is nothing that justifies moving to fully ray-traced scenes or at least tracing of triangle based data.
Using ray-tracing for certain things is an another thing entirely.
Like tracing distance fields or oct-trees for things like a good ambient occlusion, shadows, soft reflections/refractions is the way to go.

Obviously we have offline renders to look at as a goal, but as mentioned, the performance to do that is a long way away, softening raytraced shadows is expensive as hell for example (judging from the performance impact in Mental Ray atleast)
When sampling soft-shadows by tracing lots of rays per pixel for each light, it is pretty much given that it's slow.
It's even worse for soft reflections & GI where you actually have to calculate shaders for each of those 'soft' rays.


Happily, one can trace things other than polygons, like sparse octree voxels where you can pre-filter your geometry.
http://www.youtube.com/watch?v=ZI2-9UpDk6I
In this case you just sample your single ray with the filtered version of the geometry and get anti-aliased and blurred version of the shadow.

[Atomizer]

One thing I have been thinking of in regard to ray-tracing and soft shadows, cant you just use an oldschool style lightmap, then when a ray samples that pixel itll pick up the shadow along with it, without any overhead on the raytracing side? (Just a slight overhead from generating the lightmap whenever it changes)
Or cant they use shaders in the same way? Sure it might need new hardware, but to make a complete raytracing renderer viable you need new hardware anyway.
I still have my doubts that a complete raytrace engine with modern graphics expectations will ever be viable though.

[jlippo]

One thing I have been thinking of in regard to ray-tracing and soft shadows, cant you just use an oldschool style lightmap, then when a ray samples that pixel itll pick up the shadow along with it, without any overhead on the raytracing side? (Just a slight overhead from generating the lightmap whenever it changes)
Or cant they use shaders in the same way? Sure it might need new hardware, but to make a complete raytracing renderer viable you need new hardware anyway.
I still have my doubts that a complete raytrace engine with modern graphics expectations will ever be viable though.
Ofcourse one may use precomputed shadows/light with ray-tracing, just as on would on rastering engines.

Differences of rasterizing and raytracing pretty much end when you find the intersection point on geometry, you use the exact same lighting models to describe surface materials. (Phong, Bling, Strauss, Cook-Torrence... warning .PDF) (http://www.devmaster.net/articles/survey-of-shading-models/a_survey_of_shading_models.pdf)
Yes, you can shoot more rays to do all those effects on raytracing for something like radiosity, but you can do that with rasterizer as well. (altough it just isn't feasible in most of the cases. ;))

[Atomizer]

Ofcourse one may use precomputed shadows/light with ray-tracing, just as on would on rastering engines.

Differences of rasterizing and raytracing pretty much end when you find the intersection point on geometry, you use the exact same lighting models to describe surface materials. (Phong, Bling, Strauss, Cook-Torrence... warning .PDF) (http://www.devmaster.net/articles/survey-of-shading-models/a_survey_of_shading_models.pdf)
Yes, you can shoot more rays to do all those effects on raytracing for something like radiosity, but you can do that with rasterizer as well. (altough it just isn't feasible in most of the cases. ;))

True, with the tech they have now for rasterizing, I dont see the need for raytracing, I remember EVE is going to be using a lighting engine for their new "Walk in stations" that does radiosity, I feel its much more practical to work with what works for real-time games, and expand upon it