Wouldn't moving to a REYES-like system help PS3 a lot ?

[Panajev2001a]

Benefits:

1.) Compatibility with High Level Shaders written for systems such as Renderman ( PRman ), Stanford RTSL ( Sony is financing this project along with nVIDIA and ATI ) etc... this would save Sony a lot of work in the creation of the PlayStation 3 SDK and would offer developers and artists a nice and very flexible High Level solution.

2.) I would expect Sony to provide developers with the REYES-like engine included in the system's SDK: they might even open the source code to some developers if they are intentioned to customize some small part of it, but there are several possibilities.

3.) Another option would be using Renderware and other middleware engines, but likely those would not extract the peak performance out of the machine, while Sony's internally developed REYES-like engine would come reasonably closer to the goal ( they could take Stanford RTSL effort and port it to PlayStation 3 optimizing the heck out of it for the specific features of the PlayStation 3 architecture even though their software was already almost thouhgt out for another stream architecture which is the "IMAGINE" stream processor IIRC ).

4.) What if the graphics are not leaps and bounds over the Xbox 2 and its more traditional rendering pipeline ( compared to a REYES-like paradigm ) ? It does not really matter: chances are that some of the strengths of REYES ( displacement mapping, Stochastic Sampling, etc... ) would still allow the PlayStation 3 look to have a certain appeal andwhat matters the most is that the REYES-like engine will be perhaps the best way to squeeze out all the graphical performance out of the machine, but let's not forget that we would not have all that Floating-point and Integer power on the PlayStation 3 CELL based CPU ( the Broadband Engine ) for graphics alone as I see A.I. and Physics benefitting from a parallel architecture such as CELL.

5.) Developers pushing the PlayStation 3 to the limit would then tend to use a REYES-like rendering engine ( Sony;s own,Sony's own with modifications or a developers' own REYES-like renderer ) which would not be too hard to use for them and their artists, certainly easier than writing ASM code on PlayStation 2 or on NV2A ( Vertex Shaders and Pixel Shaders before Cg and Microsoft's own HLSL was basically coding at the ASM level ).

6.) Point 5.) is interesting, if Sony captures a good amount of developers to support PlayStation 3 and the sales of the machine are good enough. I am not sure the ATI chips in the GCN 2/NES 5 or Xbox 2/Xenon would be optimally pushed bya quick port of those engines ( yes, I know that by the time those consoles are out, they will have flexible GPUs capable of converting without incredible hassles Renderman Shaders and run them at real-time speeds in game scenarios, but I still do not see any of ATI's rumored chips for those two consoles as being optimized around Micro-polygons ).

7.) Developers using Renderman would already level the differences in grpahics between the two platforms, but doing a quick port of the REYES-like engine you used in your PlayStation 3 title would downplay Xbox 2's strengths, something which Sony would not exactly hate to see.

8.) This would make sense paired with what people were thinking a while ago: Sony would make it so that the code is not too well adapt to be ported to other architectures. The problem is that people thought this was due to the fact PlayStation 3 would have been hard and complex to program for. I do not think that a HLSL like Stanford's RTSL would be "hard and complex to code for" in the minds of most developers.

Disadvantages: ? You guys can fill this one .

 

[Guden Oden]

Um, you think it's realistic for developers to be able to handle such a massive paradigm shift in how software is written? I'm highly doubtful about that.

 

[London Geezer]

If the power is there, great, personally i don't think PS3 will be THAT powerful. I think polygons and textures (shaders) will be around at least until PS4. And they might never really die anyway.
And yes, asking developers to have 2 engines for a multiplatform game (one *traditional one* for X2 and N5 and a Reyes-like one for PS3) might push them to only develop for one platform. That is, unless PS3 will be OK working with polygons and textures and *old* techniques as well as a completely different method of rendering. I also think the trickiest part will be the lighting. That is what would really set PS3 apart, if something like what you say, Pana, becomes true...

 

[Phil]

would someone care to explain the difference when approaching Reyes opposed to rendering done today? Thanks in advance....

 

[Panajev2001a]

Um, you think it's realistic for developers to be able to handle such a massive paradigm shift in how software is written? I'm highly doubtful about that.

They could focus on HOS ( subdivisional surfaces and NURBS ) and let the engine subdivide the objects for them.

They could do sub-pixel perfect Displacement Mapping, they could do some of the best motion blur and AA possible, they would have a High Level Shading interface and a good amount of literature in Renderman shaders available to them.

Some parts of REYES make more sense, are more organized: when you have to run the Shaders ( Textures are still used as inputs for the shaders ) you are going to apply them on a very uniform and logical substrate:no mix of strips, fans, HOS, etc... all nice micro-polygons ( flat shaded lovable thingies ).

No Vertex + Pixel Shading: all the shading is done at the Micro-polygon level.

 

[Panajev2001a]

would someone care to explain the difference when approaching Reyes opposed to rendering done today? Thanks in advance....

http://www.tml.hut.fi/Opinnot/Tik-111.500/2002/paperit/jafar_ajdari.pdf

 

[Phil]

oh thanks a lot Pana! Will check it out!

 

[London Geezer]

Not sure why i'm asking you, Pana, but do you realistically think PS3 will be THAT powerful? Personally, i think it could reproduce some of those elements in real-time, but ultimately we'll have to see if a stripped-down version of a Reyes renderer will look as good in our eyes as a full-blown "traditional" renderer, which by the time Xbox2 will come out, will be able to produce some STUNNING images at stable framerates....

I'm a bit pessimistic, which is good cause if i'm proven wrong, i'll be even happier once PS3 comes out!

 

[Panajev2001a]

London-boy, I am not saying you will be able to take the Toy-Story 2 or the Finding Nemo's RIB files and get them to run on PlayStation 3 at 60 fps.

I think it would be worth to push a more similar amount of polygons and textures per second, but have decisively better image quality and smoothness for the camera motion. I am sure to enable a REYES-like architecture you migth have to make some compromises, but the way the patents seem to describe the ideas behind the Broadband Engine and CELL in general would appear to be very efficient at working with a REYES-style paradigm.

It would fit with the evolution you make from the PlayStation 2 design, to the GSCube R&D projects forward.

 

[London Geezer]

London-boy, I am not saying you will be able to take the Toy-Story 2 or the Finding Nemo's RIB files and get them to run on PlayStation 3 at 60 fps.

I think it would be worth to push a more similar amount of polygons and textures per second, but have decisively better image quality and smoothness for the camera motion. I am sure to enable a REYES-like architecture you migth have to make some compromises, but the way the patents seem to describe the ideas behind the Broadband Engine and CELL in general would appear to be very efficient at working with a REYES-style paradigm.

It would fit with the evolution you make from the PlayStation 2 design, to the GSCube R&D projects forward.

Yeah my thoughts exactly. Sony's been heading towards a POLYGONS POLYGONS POLYGONS approach for quite a while, ultimately i think they will definately push for a texture-less platform. We just have to see which platform it will be...

 

[jvd]

4.) What if the graphics are not leaps and bounds over the Xbox 2 and its more traditional rendering pipeline ( compared to a REYES-like paradigm ) ? It does not really matter: chances are that some of the strengths of REYES ( displacement mapping, Stochastic Sampling, etc... ) would still allow the PlayStation 3 look to have a certain appeal andwhat matters the most is that the REYES-like engine will be perhaps the best way to squeeze out all the graphical performance out of the machine, but let's not forget that we would not have all that Floating-point and Integer power on the PlayStation 3 CELL based CPU ( the Broadband Engine ) for graphics alone as I see A.I. and Physics benefitting from a parallel architecture such as CELL.

If its only able to out put graphics comparable to the xbox 2 then how in gods name will reyes help it .

Not only will there be a massive drout of reyes capable programers but the games will be behind what can be done in a more traditional way and that may stay like that through out the generation.

Then in 6 years the ps4 might be more powerfull but the advancements made in the standard graphics pipeline will once again make the move a big if .


Imho

 

[Tuttle]

Um, you think it's realistic for developers to be able to handle such a massive paradigm shift in how software is written? I'm highly doubtful about that.

What "paradigm shift" are you talking about???

Every person I know who works with computer graphics has their own personal raytracer of some sort they've been working with for years. I do on my dual G5, along with my stack of image synth books sitting right here in my active reading stack, Renderman spec, and so on.

GPUs as they exist today from ATI and NVidia are clunky hacks. Being able to jettision all that go to a 'pure' source of digital image synth would be a lot easier to code for.

What we know of the PS3 hardware indicates that it would be a dream machine for exactly what Panjev is suggesting.

Using a x86 home computer with an ATI/NVidia card as a benchmark for what is 'normal' is not a good way at looking at the world of graphics.

 

[Panajev2001a]

4.) What if the graphics are not leaps and bounds over the Xbox 2 and its more traditional rendering pipeline ( compared to a REYES-like paradigm ) ? It does not really matter: chances are that some of the strengths of REYES ( displacement mapping, Stochastic Sampling, etc... ) would still allow the PlayStation 3 look to have a certain appeal andwhat matters the most is that the REYES-like engine will be perhaps the best way to squeeze out all the graphical performance out of the machine, but let's not forget that we would not have all that Floating-point and Integer power on the PlayStation 3 CELL based CPU ( the Broadband Engine ) for graphics alone as I see A.I. and Physics benefitting from a parallel architecture such as CELL.

If its only able to out put graphics comparable to the xbox 2 then how in gods name will reyes help it .

Not only will there be a massive drout of reyes capable programers but the games will be behind what can be done in a more traditional way and that may stay like that through out the generation.

Then in 6 years the ps4 might be more powerfull but the advancements made in the standard graphics pipeline will once again make the move a big if .


Imho

First, it might be easier to code for as the idea behind REYES is very logical and uniform.

Second, there would be some advantages even if the polygon count, number of textures per second, etc.. would be similar to what Xbox 2 can do with a more traditional rendering pipeline: very good AA, very good Motion Blur, great Displacement Mapping, etc...

Also, what is the biggest problem that we will have with Hardware becoming more and more powerful, a problem peoiple ee already in the next-generation of consoles ? It's content creation.

Textures, polygonal models, environments, etc... will have to be highly complex and advanced if you want to show off your Hardware and what it can achieve: wouldn't we want to make it easier for artists to bring their data in a more straightforward way from their art packages to the actual game engine ?

Would not be easier for them to use the high level shaders you can use when you create your content right into your game without much work to be done to trasnlate these complex shaders in some other shading language or sequence of less complex shaders ?

Since we are going to be facing the problems that off-line CG faced few years ago, why should not we take advantage of the tools and knowledge they have already gathered ?

 

[V3]

If it'll help to produce something like this, why the hell not 8)
http://www.virtualcinematography.org

ATI compared their current solution to this in one of their GDC courses, and it make them looks really bad, even though its probably impressive compare to any current real time solution.

 

[London Geezer]

Just thinking that one day games will look like that made me skip a heart beat... Of course, far from now and PS3, but one day we will reach that...

 

[_phil_]

Using a x86 home computer with an ATI/NVidia card as a benchmark for what is 'normal' is not a good way at looking at the world of graphics.

100% agree with that ,perception has been so bended by the 7 past years of the "mighty combo" of eyecandy paradigm....
What i missed the more is certainly software rendering and its freedom and control.

 

[Simon F]

Yeah my thoughts exactly. Sony's been heading towards a POLYGONS POLYGONS POLYGONS approach for quite a while, ultimately i think they will definately push for a texture-less platform. We just have to see which platform it will be...

IIRC, I've read that the typical (Pixar) Movie Renderman frames may also access 100s of Megabytes of texture data.

Can't see textures going away anytime in the next few years.

 

[Guden Oden]

What "paradigm shift" are you talking about???

Um, traditional textured polygon graphics to Pana's reyes micropoly vision perhaps?

Every person I know who works with computer graphics has their own personal raytracer of some sort they've been working with for years.

You probably work with a very small sub-set of computer graphics workers.

GPUs as they exist today from ATI and NVidia are clunky hacks. Being able to jettision all that go to a 'pure' source of digital image synth would be a lot easier to code for.

But everybody would have to start over from scratch. Not only make entirely new dev-tools, but also re-learn how to build engines and contents that runs fast and effectively on the hardware in question, and if PS3 is the only machine that works this way it would be a serious drawback, particulary for small developers. They couldn't share code or graphics contents of PS3 games with other platforms.

You could probably program a reyes-based PS3 to run as a traditional polygon renderer too so you wouldn't have to build two completely different versions of the same cross-platform title, but then what would be the point of moving the PS3 to a different type of architecture?

What we know of the PS3 hardware indicates that it would be a dream machine for exactly what Panjev is suggesting.

Problem is, we don't ACTUALLY know anything about PS3...! It's all smoke and mirrors and bad guesses right now.

Using a x86 home computer with an ATI/NVidia card as a benchmark for what is 'normal' is not a good way at looking at the world of graphics.

Well, normal or not, but the clunky polygon rendering approach has proven itself to be very effective despite whatever drawbacks exist. People are used to working in that environment, and re-training them is going to cost a lot of time and money.

 

[London Geezer]

What "paradigm shift" are you talking about???

Um, traditional textured polygon graphics to Pana's reyes micropoly vision perhaps?

Every person I know who works with computer graphics has their own personal raytracer of some sort they've been working with for years.

You probably work with a very small sub-set of computer graphics workers.

GPUs as they exist today from ATI and NVidia are clunky hacks. Being able to jettision all that go to a 'pure' source of digital image synth would be a lot easier to code for.

But everybody would have to start over from scratch. Not only make entirely new dev-tools, but also re-learn how to build engines and contents that runs fast and effectively on the hardware in question, and if PS3 is the only machine that works this way it would be a serious drawback, particulary for small developers. They couldn't share code or graphics contents of PS3 games with other platforms.

You could probably program a reyes-based PS3 to run as a traditional polygon renderer too so you wouldn't have to build two completely different versions of the same cross-platform title, but then what would be the point of moving the PS3 to a different type of architecture?

What we know of the PS3 hardware indicates that it would be a dream machine for exactly what Panjev is suggesting.

Problem is, we don't ACTUALLY know anything about PS3...! It's all smoke and mirrors and bad guesses right now.

Using a x86 home computer with an ATI/NVidia card as a benchmark for what is 'normal' is not a good way at looking at the world of graphics.

Well, normal or not, but the clunky polygon rendering approach has proven itself to be very effective despite whatever drawbacks exist. People are used to working in that environment, and re-training them is going to cost a lot of time and money.

Guden, i think he was (quite rightly so) pointing out that 3D models used in games are *often* created using raytracers to begin with, then scaled back from hudreds of thousands of polygons, to models that are useable in today's games. So, having a platform that lets you use your original raytraced model would be useful, and not a paradigm shift at all. There would be no re-training to do, since to get to gaming development you need to know a lot about raytracers to create your models in the first place.
It would be great for me for example, i have experience with Maya and i could very easily just get into game development, something that i couldn't do nowadays without some sort of additional training.

But i agree, it's all smoke and mirrors...

 

[GameCat]

I don't think the difference between the REYES architechtuire and modern hardware is that great. You can already do a lot of Renderman sahders in hardware (although the resource limits on current gen puts a stop to most longer shaders, this wion't be the case with next gen). The micro polygon architechture essentially enables per pixel displacement mapping, robust handling of transparent surfaces, and high quality AA, both temporal (motion blur) and spatial (regular AA and depth of field). None of this is that hard to do even on current hardware except transparent surfaces. And I honestly don't think that will provide such an immense difference to developers.

But honestly, I'm no REYES expert, is there anything more specific you guys think will be better in game graphics with such an architechture?