Clearspeed announces CELL-like processor

Pixel Shading in the future should come as a geometry compression mechanism rathern than enhance graphics quality in itself as when triangle size approaches 1-10 pixels the difference between per-pixel lighting and vertex lighting is less noticeable.

Of course you do not want to tesselleate that finely a flat wall just to get lightign right: unless you go for a REYES like approach which has its own advantages.
 
Whatever issues PS3 may have, I doubt a lack of FP power will be one of them. It could be a fraction of a teraflop and still have enormous power (imagine 500 gigaflops, like that's NOT fast??). Some more significant issues:

- good compilers/developer tools (although unfriendly hardware hasn't hurt PS2 at all)
- image quality (good AA is needed, especially at HDTV resolution)
- vertex lighting, or realistic lighting? (cg quality?)
- profficiency at running shaders
- ability to texture (displacement maps, procedural noise)
- better animation/physics/collision (a software issue, the resources for it are there)

PS3 can't just offer PS2 games with 1000x more polygons, an overall increase in realism is needed.



--------------------------------------------------------------------------------

Your right Josiah, It can't be a PS2 where the main difference was just pure polygon power.

However, I am sure that Sony has learned their mistakes(PSP is showing this) and PS3 will be chock full of features and not just the ability to push a billion polygons around.

On the Shader(effects) side of things Sony's GPU will have it's own APU's, meaning that developers can just program their own effects in(think per pixel lighting) without them having to be stuck with what's locked in hardware.

Kutaragi has also said that there will be developers libraries, and a API for ps3; I am certain that there was no graphics API for ps2.
Click to expand...



two exellent posts. i agree with both
 
Panajev2001a said:
Pixel Shading in the future should come as a geometry compression mechanism rathern than enhance graphics quality in itself as when triangle size approaches 1-10 pixels the difference between per-pixel lighting and vertex lighting is less noticeable.

Of course you do not want to tesselleate that finely a flat wall just to get lightign right: unless you go for a REYES like approach which has its own advantages.
Click to expand...

ah, but even a REYES approach uses pixel shaders (much more so than any 3D game, actually). what's really needed is a more accurate simulation of the nuances of how light behaves (ie subsurface scattering).
 
Josiah said:
Panajev2001a said:
Pixel Shading in the future should come as a geometry compression mechanism rathern than enhance graphics quality in itself as when triangle size approaches 1-10 pixels the difference between per-pixel lighting and vertex lighting is less noticeable.

Of course you do not want to tesselleate that finely a flat wall just to get lightign right: unless you go for a REYES like approach which has its own advantages.
Click to expand...

ah, but even a REYES approach uses pixel shaders (much more so than any 3D game, actually). what's really needed is a more accurate simulation of the nuances of how light behaves (ie subsurface scattering).
Click to expand...

Sub-surface scattering is a very nice idea indeed: it can really enhance the look of certain 3D models ( which try to simulate certain materials ).

About REYES, well ( not to be a pain, but... ) REYES basically only does vertex shading as the shading phase is done at the micro-polygon level ( which are still flat-shaded, less than a pixel in size quadrilaterals ) the only big part that is done at the pixel level is stochastic AA ( edge AA, motion blur, Depth of Field, etc... ), but I agree that for all intent and purposes we have vertex shading and pixel shading squashed in one stage.

I was only trying to separate what is needed for good lighting with vertex lighting and the geometry compression and a bit lower demands on T&L pixel lighting ( DOT3 based ) brings in a way ( need to transform less polygons just to give lighting a realistic look, you might have already passed the point in which the 3D model is already quite smooth ).

I was not trying to say that an approach based on micro-polygons or ultra-finely tessellated geometry would not require the use of shaders, but just that was the path you would have to get on if you wanted to push lighting at the vertex level.
 
What I saw in regards to Physics in HL2 was already incredibly amazing, done with a p4 1.7

Throw a body against a wall made of wood and watch the wood wall recoil back and individual pieces take the force of the impact differently. Not even a gimmick in the way it's done, it's all totally convincing and real.

Some of the physics in HL2 are so life like it's almost scary, me and my friends could spend hours messing around with that one e3 scene(room with materials and huge pool). Incredible.

Although some of the physics in the link above are beyond anything I have ever seen, something beyond amazing; truely insane.
 
london-boy said:
:oops: oh my god those physics demos look amazing...... just.... amazing...
Click to expand...

I know - so I don't know why people are worried about how to use the FP power - there will always be a way to use it.

No bump-mapping in those physics demos - yet infinitely more realistic. With the kind of power the next-generation systems will provide, I think that "realism" will come through more accurate physical models of object behavior, lighting, and particle effects, rather than elaborate tricks. And it will make the game artists' and animators' jobs a lot easier - potentially significant as game budgets rise with each generation.
 
No bump-mapping in those physics demos - yet infinitely more realistic. With the kind of power the next-generation systems will provide, I think that "realism" will come through more accurate physical models of object behavior, lighting, and particle effects, rather than elaborate tricks.
Click to expand...

Heh of course there is no bump mapping, no need! Use polygons for your bumps.

I hate Bump mapping anyway, atleast when used in a bad way like that Abducted game. Halo 2 uses it INTELLIGENTLY though, which is fine. These physics demos prove that you don't need bumps to create realistic visuals.

One day the power will be there where you don't have to use hacks to simulate something(bumps on a wall) but for now.. the tricks will do.
 
Having worked for around 3 weeks for Toshiba on a project directly relating to Cell(getting it into a device), and hearing some of the other engineers who were actually working on Cell sillicon speak; I can tell you that the power Sony is planning to shove into PS3 is of gigantic, no, almost unreal in terms of power.

Your talking a few huge chips going into this machine, in which the initial losses are going to be gigantic. Sony obviously must know this, but hey, Sony has always been going for huge specs.
 
Tokyo! said:
Having worked for around 3 weeks for Toshiba on a project directly relating to Cell(getting it into a device), and hearing some of the other engineers who were actually working on Cell sillicon speak; I can tell you that the power Sony is planning to shove into PS3 is of gigantic, no, almost unreal in terms of power.

Your talking a few huge chips going into this machine, in which the initial losses are going to be gigantic. Sony obviously must know this, but hey, Sony has always been going for huge specs.
Click to expand...

This would explain why they are not dropping PlayStation 2's price even further: they are tying to make cash and tons of it to prepare for PlayStation 3's launch.

This will take some accounting magic though: there has to be a way of putting the extra cash they make in the books later to balance PlayStation 3 losses in its first year out.

What you tell us could be understood if you look at the efforts SCE and Toshiba are putting in the development of their next manufacturing process ( 45 nm SOI with those sweet capacitor-less DRAM cells ).

They have said over and over how they plan to have the 65 nm lines upgraded to 45 nm technology and they have said this before they started building Oita #2.

The CELL chips in PlayStation 3 are going to be pretty nice :)

What worries me still is the software, they have LOTS of work to do to make PlayStation 3 SDKs ( APIs, tools, etc... ).

This happened at GDC2003:

A developer from Sony asked if anyone had any experience updating objects on an SMP system. No one did, but if the PlayStation 3 really does have 72 processors, I suppose we're all going to learn.
Click to expand...

I can imagine the face on Sny's guy... lots of work is ahead for the software guys.
 
This would explain why they are not dropping PlayStation 2's price even further: they are tying to make cash and tons of it to prepare for PlayStation 3's launch.
Click to expand...

As I mentioned to you earlier, this is actually what I was thinking aswell. They make as much money as they can off of PS2 now and bank some of it to sort of normalize PS3's gigantic losses from PS3's first year.

If it's one thing for certain it's this, Sony wants the performance advantage with PS3, even if just in raw spec. Everything that has been done so far shows this, including the whole rush down to the 45 nm node. This shows that they plan to take huge losses in which they don't want to bear for long.

Software still has a bit more to go, although IBM is at work on this.
 
...

Having worked for around 3 weeks for Toshiba on a project directly relating to Cell(getting it into a device), and hearing some of the other engineers who were actually working on Cell sillicon speak; I can tell you that the power Sony is planning to shove into PS3 is of gigantic, no, almost unreal in terms of power.
Click to expand...
I hope you will be staying around on the day Kutaragi Ken makes the first public demonstration....
 
Tokyo! said:
Having worked for around 3 weeks for Toshiba on a project directly relating to Cell(getting it into a device), and hearing some of the other engineers who were actually working on Cell sillicon speak; I can tell you that the power Sony is planning to shove into PS3 is of gigantic, no, almost unreal in terms of power.

Your talking a few huge chips going into this machine, in which the initial losses are going to be gigantic. Sony obviously must know this, but hey, Sony has always been going for huge specs.
Click to expand...

Sweet. Good to hear.

No offense to you personally, but I'm a skeptic when it comes to information that has no link to a credible source. Obviously, with stuff like this, it is impossible to get a link. Is there anyway you can prove that you work for Toshiba without violating whatever security procedures Toshiba had you sign on to? Perhaps tell us more about your work?

And then, please give us as much info as you can... :)
 
...

Come on, does he sound credible to you? "Worked on CELL for three weeks?:" Yea, and I figured out what it was three years ago...
 
Re: ...

DeadmeatGA said:
Come on, does he sound credible to you? "Worked on CELL for three weeks?:" Yea, and I figured out what it was three years ago...
Click to expand...

Didn't you intern (field insignificant) during your summers while still an undergrad? I can't think of a person in the sciences with any ambition who did not... atleast once.

With all due respect, I don't have faith in you figuring out much.
 
A rumour is going around that THQ are going to start working on XBOX2 and Sony PS3 games in the next two weeks. Apparently under emulation provided by Microsoft and Sony. THQ cite the need to start so early on games on the fact that Microsoft or Sony may release their consoles earlier rather than later (2005 rather than 2006).

If this is true (and the source was someone from THQ) then there should be announcements or leaks from other developers as well in the next month or so.

Edit: I found a link here: http://www.cube-europe.com/news.php?nid=5610 which has a link to a Reuters article. It must be true then eh? ;)
 
Re: ...

DeadmeatGA said:
Come on, does he sound credible to you? "Worked on CELL for three weeks?:" Yea, and I figured out what it was three years ago...
Click to expand...
The question is, do you sound credible when you bash SCE 24/7 and products they are working on that are still 2 years away ( like you also did for the PlayStation 2 ) ?
 
Panajev2001a said:
Yea, and I figured out what it was three years ago...
Click to expand...

You also moved to somethign much better and which is much more practical...

Can I see the patent you filed about this wonderful new tech ?
Click to expand...

Sony beat him to the patent office.
 
You must log in or register to reply here.

Similar threads

D
Game Economics, Graphics and AI ...
Replies
0
Views
380
DavidGraham
D
Cyan
NPU (Neural Processing Unit). AI Hardware Acceleration discussion.
2
Replies
34
Views
4K
pcchen
P
M
Interesting old GPU architecture
Replies
17
Views
3K
Ailuros
Ailuros
Cyan
Play on your PC as if it was a console, on Linux, Windows, etc.
Replies
1
Views
1K
Cyan
Cyan
S
Apple Dynamic Caching on M3 GPU
Replies
5
Views
3K
Lurkmass
L
Back
Top