Caustic Graphics develops ray tracing at gaming speeds
By Joseph L. Flatley 
posted Mar 13th 2009 12:57PM
posted Mar 13th 2009 12:57PM
The ray-tracing technique for generating 3D images is an extremely processor-intensive proposition that doesn't lend itself to gaming (or real-time anything, for that matter), but all of that may be about to change. Caustic Graphics, a San Francisco-based start-up founded by former Apple employee James McCombe, claims that its dedicated ray tracing card will produce photorealistic graphics at a rate 20 times faster than products currently available. And if that weren't enough, their new CausticRT platform has the potential to perform the operations a whopping 200 times faster -- possibly fast enough for a photorealistic gaming platform. The second generation technology should be unleashed sometime in mid-2010, while the CausticOne card will get its official announcement on April 23 of this year.
[Via Business 2.0 Press]
[Via Business 2.0 Press]
On the front page of their website, why did they rate their little graphic with an ESRB rating?
Because they CAN?
Rated (R)? Lol
Sweet, I was just thinking about how today's games lack graphical realism.
I'm not thinking games, I'm thinking what a difference this'll make to my rendering speeds, definate buy if it performs as promised.
+1 for your user pic :)
RH FTW!
I wonder how this would compete with Intel's Larrabee
Its a completely different concept. Comparing this with larrabee is comparing apples to oranges. The caustic hardware is actually specifically designed for ray-tracing data pre-processing not the processing itself, it would actually send the data off to the GPU/CPU for the actual processing.
Bring it... something has to take down DirectX.
What the hell is wrong with DirectX?
Please explain because I thought Direct X was great. Producing better
graphics with less powerful machines..
DirectX is good, but one can only hope something comes in and beats the crap out of it for the sake of evolution. If something threatens directX, graphics could leap forwards because of some real competition.
OpenGL is just copying Direct3D. DirectX is really where the innovation is.
You realize OpenGL has been around for longer than DirectX, right?
Penguin you realize you are saying crap don't you ?
DirectX was the best marketing product ever invented by Microsoft so yes in a way it IS great.
On the other hand it has been the (indirect) reason for Linux poor GPU support for a long time...
Now everyone will be able to run Crysis.
Crysis is DirectX. This is OpenGL.
They are probably using Voxels like everyone else is starting to do, but don't want to let out the "secret"....
http://www.beyond3d.com/content/articles/94
Have a read. It's enlightening. As a 3D modeler, this news is a mixed bag of yay and neigh. There's a VAST difference between ray-tracing and GOOD ray-tracing.
W-O-W.. If this does what it says it does... amazing. Even extremely simple ray tracing tends to take massive amounts of time. This would revolutionize gaming.
Ray Tracing is also easier to program.
Next Next Gen: F'real This Time
@ Gibson: You fail. It is markedly harder to program for ray tracing, and EVERYBODY (including Intel and Nvidia) has stated such. Go get a life.
MikeC, I think you misunderstood. Ray-tracing programming is actually significantly easier to program and conceptualize. What is hard is making it faster. You have to get into spacial structures and concurrency algorithms to increase speed. There are few techniques now that really speed up ray-tracing. The latest improvement to ray-tracing is GPGPU architectures like NVIDIA's Tesla. To my understanding Caustic's hardware will supplement OpenCL/CUDA/CTM and make the process that much faster. Its all really interesting, I am currently working on ray-tracing with CUDA, and it all looks very promising. Especially with Caustic's plan, I feel that ray-tracing will be widely used video games within 5 years.
If you still don't believe that ray-tracing is easier, don't take my word for it, do some research.
when i read raytracing, i imagine those 3D animations of shiny superreflective spheres bouncing around, in the early 90s shows....
they love chrome back then...
but without them, my Trapper Keeper folders would look so much less psychedelic.
In old raytracing stuff, you always see perfect sphere because they are very simple to render at what is essentially an infinite number of "polygons" because of the nature of raytracing (and other similar rendering techniques)
Not sure, but I'd like to know if this has any impact on how we do collision detection. With normal rasterization, you have to calculate intersections yourself. However, with hardware accelerated raytracing, I would imagine doing ray/poly hit testing would be blazingly fast. I wonder if collision detection algorithms would degenerate into the exact same code as the rendering engine.
Yeah, fast enough for games if that game is "guess the answer" or something. I'm sure this will be cool, but as games progress, the detail levels get insane, and raytracing gets even less of an option. Plus, the current shader tech is really good - glowing, bumps, etc. This would also need to be incorporated in before it even was remotely applicable.
For the Voxel unbelievers, id's next game "rage" is voxels for everything except cars and people: http://www.pcper.com/article.php?aid=532
Almost...
Rage is based on the id tech 5 engine, id tech 6 will use voxels in future titles.
MS should use the 2nd generation one in their next Xbox!
I seem to remember a company outside of the normal ATI, Nvidia, and Intel trifecta they promised big things. I wonder whatever happened to them.
This sounds great, but from everything I have read they don't have an actual product to deliver yet. This sounds just like those old Voodoo cards back in the 3Dfx days. They promise to do amazing things only to be outperformed by the much cheaper offerings from the big boys.
If this comes to real fruition it could be amazing, just take an example from 3Dfx don't aim to high with your product and make claims you can't deliver on. Don't focus solely on the high-end ignoring the budget gamer. I will be watching this develop with great interest.
fingers crossed
Those 3dfx cards did all of the things they said they did. The issue is their raw rendering speed and cost were not up to snuff. Imagine if ATI released a card now that had an amazingly advanced feature set for new effects that no other card had, but that it costa lot more than NVIDIA's offering while being slower overall when performing traditional rendering. No one is going to buy the ATI card based on what it could potentially do if developers supported it and take a hit to the current and upcoming games when they could just get the faster NVIDIA card which is faster right overall for the games of the time. There was more to it, but thus was a big thing.
It is hard to get an industry to shift paradigms drastically at the drop of a hat. It requires a large initial expenditure for what might be a minimal improvement. It is one of the reasons backwards compatibility keeps being touted while hybrid platforms get pushed.
So can this card do real-time caustics rendering as well? It IS from Caustic Graphics! If a card can do raytrace, caustics, and radiosity all in one pass in real-time, that'd be the day we see real-time realistic graphics gaming!!!
CausticOne, the Physx of 2010! Well we know what happened there. To recap, Physx trotted out their PPU and the company was quickly snatched up by Nvidia (who incorporated the tech in their GPUs). Oh and there was also that thing about games needing to support it. So Caustic don't be a bunch of CTs. Partner with one of the big dogs (to incorporate it in some new fangled GPU) so that it can become standard fare quickly. Or the other option is to party with Sony and the Cell processor for some PS4 or 5 uber ray tracing game action? Who knows, exclusivity to your tech might make you the "king makers" of the next gen console warz...
PS Like that last sentence do ya? I see you all smiling ;^)...
Anyone else think Apple might buy this up? Apple might get seriously into gaming.
Maybe if they bring out a mid-range tower mac, and compatible graphics cards that aren't overprice and underpowered, then I'll think about it.
and do what with it exactly? Apple gets their graphics cards from ATI and Nvidia, if they snatched it up they'd have to make their own graphics cards, something I personally doubt they'd do. Not to mention One of the big boys in this area(the aforementioned ATI or Nvidia) would probably be willing to pay a lot more than apple for it and would have more to gain by making it pc-only than mac-only.
Yeah when pigs fly.
I'll never understand why Apple goes so far out of its way to ignore the gaming market. Here you have a company who sells computers to consumers by marketing it as "Much more fun to use than your WORK computer". The company is run by a guy who also started a CARTOON company, and yet they don't sell a gaming computer (not under $3k anyway), and don't really reach out to gaming companies.
It seems like Games & Apple ought to be a natural fit.
Yeah sure !
Hardcore gamerz have always been HUGE Apple fans...
Look at all these cheap, easily modded, overpowered devices they put to the world : Macbooks,and ehrr... how do you guys call the big plastic box with 3years old GPU that you pay $3k again ?
Oh and the Mac mini ! definitely the gaming device !
You're just so far from a good idea I can't even tell...
No, dude. If Apple decides to stick their little fingers into gaming, I'm leaving, and you should too.
There's only so much you can do with a particular amount of processing power. We will see realtime raytracing eventually, but it's not going to suddenly appear with the launch of some magic '20 times faster' product. And if they have an engine that can do something 200 times faster, more than likely it is using massive simplfications or 'tricks', and hence its not really raytracing anyway.
Wrong! It is gonna suddenly appears with this kind of people i.e. people who decide to put financial power to develop hardware specifically for this type of application.
Once this kind of possibilities are offered to game developers they will make games using the technology.
The other way round is much much less likely....
The other Option is to wait for CPU to get 1000 times more powerfull so they can do real-time ray tracing while decoding TRULYREALMAXHD (I think this will be the next HD standart after HDready and Full HD I don't see how they could call it another way) And running Crysis 12. Of course not to mention this CPU will use 10 times more power than the dedicated chip the guys has been developing..
So you wait for this CPU to come and I'll go check out what these guys have to offer ok ? :)
The problem with current GPUs is that they don't accelerate Ray tracing algorithms. They are built around general floating point calculations. If someone concentrated on accelerating the standard algorithms behind ray tracing then we might see some quick movement.
Exactly, it has to be done by someone else, because neither ATI/AMD nor Nvidia (or Intel LoL) is gonna go this way.
Ultimately, this new chips might be integrated to Nvidia or ATI chipset of course..
After reading all the comments, it almost sounds like you are pretty happy paying $400 for little upgrades, whatever happened to nnovation and bleeding edge? Even the article upthere explains what raytracing is and how difficult to render by current machines.
We need good competition, Nvidia even manages to sell faulty chips because they know people will still bend over for them. Hopefully, Nvidia has to pay through the nose once it tries to buy this company.
I don't give a crap for an improvement in video games after seeing killzone I think we're doing just fine. An improvement in render time for the smaller animation and visual effects studios would be awesome though. If I were this guy from Caustic, I would watch my back because he's about to put a whole lot of render farms out of some serious business ( rendering frames for hire)
Hey everyone,
I personally think there is potential here, but the company also has a massive task ahead of its self with actually getting consumers beyond the hardcore gamers to actually purchase it....plus with intensifying competition, Nvidia who may actually release a similar product, and u all know how hard it would be to compete with nvidia....
That's what this is targeted to. They state on the site that it's intended for workstations and 3D modeling applications. Engadget added the "gaming vibe" to it.
I think you missed it. This is not for gamers, yet. They are marketing this towards the film, architecture and prototyping industries. I can tell you that in the film industry anything to reduce render time will be used by everyone in the industry.
200x isn't anywhere close to what gaming will need. Don't get too excited.
And - what's more important is - your mileage will vary. 200x is probably under controlled demo environments.
Does it play Doom??? if it does, how play players can it render simultaneously???
Amazing, and revolutionary, if true.
HOWEVER - I thought the industry learned it's lesson when all the following dedicated singular-purpose expansion processors failed: a) X-Fi (thinking of X-RAM here), b) Ageia (before they got swallowed by nVidia, c) AISeek Intia (artificial intelligence)
Chicken and egg problem - developers won't develop for the card if no one has the card, and consumers won't buy the card if there are no games for it.
I think your right, thats why the best choice is for them to partner themselves to a major company and intergrate their technology with current cards.
True and maybe this will be a epic fail on the commercial point of view, but it might become just a feature of a new generation of "GPU" which would implement traditional GPUs with this kind of architecture, all-in-one.
My question is, why in hell is it based on OpenGL? If you want speed, you use anything (ANYTHING) but OpenGL.
Are you an idiot?
That's a serious question. OpenGL is not and never has been slow.
Telanis.... I don't know what to say. Let's just leave it at "fail" (I'm being nice even).
If you want to correct answer: OpenGL allows companies to write extensions (Direct3D doesn't because microsoft wants a standard interface, not a bunch of different styles used by a bunch of different companies)
My favorite comments are the ones where people who have no clue what they are talking about are so sure of themselves.
Well anything but OpenGL is direct3D...
And you are not gonna build a new technology based on Direct3D, because you would be limited by the possibilities of Direct3D, whereas OpenGL is Open source so you basically can improve it up to your needs.
Where are the god damned demo videos?
weird, on the website, they have a pic of a locomotive, taken from
http://www.3dlinks.com/gallery_fullview.cfm?id=4134
that was posted in 2002.
I thought that pic looked familiar...
soon this company will be acquired by NVIDIA or ATI. much like physix....
Can't wait for 3d apps to take advantage of this.
I for one found the whole realtime raytracing for games kinda pointless, since I never found the lack of realism in game came from lack of reflection(I really thought they did pretty good in that), which is exactly what raytrace is good at, but the lack of indirect illumination, which is the essential reason for 3d to look 3d. Anyone who used a 3d app, or has some basic traditional art training should know that it is the complexity of ambiant lighting that gives everyday objects volume for them to look realistic, but what raytrace gives you are mainly glossy reflection or refrection, which is hardly the point. Currently 3d software use photon tracing, final gathering or ambiant occlusion to get indirect illumination, but none of them doesn't require massive amount of precomputed data and assuming a static scene. Anyone solve that would be the time to get really excited.