PS3: Phil Harrison addresses the real-time vs CGI issue

[Josh378]

The continued vagueness of the realtime answers leads me to believe that KZ2 was pre-rendered, but more of the CG variety than the in-game variety. I'd love to be wrong on that. I REALLY wish someone would just ask the direct question. Of all the demos, KZ2 was the best. Phil mentioned Motor Storm using unusable camera angles (duh...shakey cam would make it undriveable) but can someone just ask, "Was the Killzone2 demo rendered to spec with in-game assets?" A simple question that will end the debate. KZ2 is the demo that brought the house down.

PlayStation 3, I think, is going to be cheaper to develop for than the corresponding period of PS2 development. I know that's a fairly contentious statement to make, but there's a very good reason for that. When we announced the collaboration with NVIDIA, we just talked about them making a chip - actually, they don't make anything, they're a designer, and the RSX contains an NVIDIA-designed part, which gives us fantastic GPU capabilities. But what it also gives us, and this is actually the most important bit of it, is all the toolchain and CG pipeline that comes with it, which is a very well understood development pipeline in the PC community - and, yes, in the Xbox community, frankly.

This quote I found most interesting. So RSX is a Sony chip still with NVidia IPs in it, right? Is there any hope for Chaperone still becoming a reality? PEACE.

What is chaperone?

-Josh378

 

[London Geezer]

What is chaperone?

-Josh378

There was a rumour running around about there being a 3rd chip dedicated to "reflections and shadows". Just a rumour though.

 

[Luminescent]

When IGN said the desert demo was running on 2 cells, did they mean 2 entire cell processers (PPU+7/8SPUs) or did they mean it used only 2 SPUs?

Hopefully it's the later.

 

[Phil]

Their definition of 1 CELL is 1 PPE + 8 SPEs. They state so at the beginning of the article.

 

[one]

I'm not trying to debate against you (just saying in advance) but did you see the demo? Or was the way that they described the demo give you the impression that its not possible.

I just trying to get an understanding about this.

Yes the demo is widely available and it's just a landscape fly-over. There would be no reason to raytrace anything on it, since there are no reflections. Maybe the shadows could be raytraced, but what would the point be? Besides, if any raytracing were taking place, we'd have heard from Phil Harrison.

It's just a rather limited demo, view distance is not even that great, even compared to current generation games.

The big thing about it is that it was created in realtime by Cell (no RSX intervention) from a height map and a colour map (i think).

Actually the demo was done solely on Cell by raycasting.

 

[Shifty Geezer]

Raytracing is a processor intensive way of rendering. Demonstrating Cell's maths ability, raytracing is a good visual demo. Just because there's no obvious advantage (no visible reflections for example) doesn't mean it wasn't ray-traced.

AFAIK a ray-tracing engine is also fairly simple to write, far simpler than alternative approaches.

 

[LeGreg]

There were no triangle
so it was probably not a rasterization..

Raytracing is not all about reflections, anyway.

This image is raytraced (with photon mapping), but where are the reflections ?
http://graphics.ucsd.edu/~henrik/images/imgs/hogum3pm_clouds.jpg

 

[one]

Interesting article, I think that he was dodging the real time issue a bit, but still that was encouraging. He didn't say anything about that terrain demo running on 2 cells though

All the server blades of CELL have 2 CELLs, but it's unclear if the workload of the demo really required 2 of them or not

 

[DudeMiester]

The main advantage of raytracing is that is scales O(log n) with scene complexity (e.g. poly count), rather then O with rasterisation. Not because you can do pretty reflections, even though that's another great feature.

 

[Titanio]

The Doc Ock head - the Alfred Molina head - is actually more of a Cell demo than it is a graphics demo, because we're calculating hugely complicated light sources in real-time on the Cell, even to the point where we calculate the angle at which light enters the skin, the way that the light is then coloured by your blood, and the way that it is then reflected back out. It's something called transmission. Skin is hugely complicated - if I put my finger over a light, for example, you can see that the light is coming through my skin. We were simulating that - emulating, simulating, kind of a fine line - we were simulating that on the Doc Ock head demo.

So that's really pushing the Cell rather than the graphics chipset?

Yeah. Those are really hardcore maths problems which the Cell is really good at solving.


It's not just the RSX that drives the graphical quality, then - the Cell can also really be used to improve the graphics.

Well, I'll give you a couple of other examples. The terrain rendering demo that was done by STI, which is the people who developed the Cell, doesn't use the graphics chip at all. That 3D landscape was generated in real-time from two input data sources and a software renderer running on the Cell created the final image. All that it does is output as a bitmap straight to the video hardware - it doesn't even create a single polygon, there's no concept of a polygon in that demo.

I found this stuff to be more interesting. So there may be a closer link between cell and the gpu for graphics tasks? Or at least, Cell could help compute some things dynamically that usually the GPU would access as precomputed maps etc, because they're so computationally expensive?

 

[Inane_Dork]

In games, you'd use the RSX to color skin, not the Cell.
So I'm not sure what the point of that was.

 

[Shifty Geezer]

On hearing that, I thought this stuff wouldn't happen on RSX. But then the nVidia guy took the stage and showed SSS running off the RSX with translucent hand things. I don't know what level of HDR-image derived the shading is possible on RSX, and whether the Cell could provide a measure of this. I would assume Cell was running close to the metal on that Doc Oc demo - tech demo's usually push the hardware to make the demo as impressive as possible, so I wouldn't have thought in a game situation Cell wuld offering much support, depending on the game. I huess in something like Tekken, two characters, Cell could do the lighting if needed.

 

[mckmas8808]

So are you guys saying that the cell will do little to help the graphics in a normal game? I find that hard to believe. I think the reason that they showed the Doc Ock demo, Getaway demo, terrian demo, and exploding gas station demo was to show gamers and really developers what the cell can do with helping graphics.

I think the normal everyday developer way of thinking would be like this....

In games, you'd use the RSX to color skin, not the Cell

so I wouldn't have thought in a game situation Cell wuld offering much support, depending on the game.

See I think Sony was trying to open the minds and give developers an idea of what the cell can do. They want devs to think outside the box. Thats why the games that look the best at the conference were in house. The Sony developers are already know what the cell is capable of doing.

 

[Titanio]

In games, you'd use the RSX to color skin, not the Cell.
So I'm not sure what the point of that was.

I mean, using Cell to feed the pixel shaders more dynamic data. Often with many more complex effects you'd feed it static, precomputed data because it's too much to calculate every frame. But if Cell can handle some of those harder calculations and pass the results to the GPU for final rendering, it could open some interesting doors. With the Doc Oc demo, they seemed to be doing some "heavy" lighting calculations on the CPU and passing that data to the GPU for its use there - that data was probably input for the pixel shaders.

There's a lot of headroom on Cell for "extra" stuff like that I think. Not sure why you couldn't do stuff like this in games (?) (I don't mean specfically the techniques used in the Doc Oc demo etc. but just generally leveraging some CPU power for "better" data input to the GPU, be it for lighting, procedural stuff, whatever)

SSS running off the RSX

SSS?

 

[Shifty Geezer]

I think the reason that they showed the Doc Ock demo, Getaway demo, terrian demo, and exploding gas station demo was to show gamers and really developers what the cell can do with helping graphics.

I think the Cell demos were to show how powerful the Cell is ina way that can be communicated and be impressive. They could say 'we calculated Pi to 140,000 decimal places in one second' or something but that doesn't have oomph. By showing a software renderer, things devs know are complicated, running in realtime, they exhibit in a visual communication its capabilities.

Though I don't deny Cell will have it's uses (there's only so much physics and AI you can add!) primarily I don't imagine it'll be a graphics helper. If it IS, and can ray-trace in real time, this begs the question why Sony didn't just go with a dual Cell config and leave nVidia out of the picture? Obviously RSX is bringing something to the grapihcs table that Cell can't compete with.

 

[jvd]

Rsx ? anistropic filtering , fsaa , texture compresion , framebuffer compresion , hidden surface removal . alot of hardwired features , pixel and vertex shaders .


Basicly all of nvidia's insight into the market of gpus since they were formed. Which is alot more than sony has . A cell would be nice but even a dual cell set up would be stomped by a x360 in graphics

 

[Coola]

Now THAT is cool. Not many people have 2 HDTVs but most of the people owning an HDTV also own another "normal" TV.

I'm planning on buying a new big HDTV set next year. Hopefully, [PC] LCD's price would drop down around the same time so I could maybe pick a cheap 15+ inch to go along side the TV. Great for messages and such (defending on the software).

cool. I'd like to see that video interview. is it online avaliable for d/l or stream ?

I think he's talking about the G4 interview? Because that's what I saw last night (though I can't remember every words). And I think it's already have it online:
http://www.g4tv.com/e32005/videos/index.html

- Z

or the teamxbox video interview

http://xboxmovies.teamxbox.com/xbox/1957/J-Allard-Video-Interview/

 

[mckmas8808]

Basicly all of nvidia's insight into the market of gpus since they were formed. Which is alot more than sony has . A cell would be nice but even a dual cell set up would be stomped by a x360 in graphics

Perfect answer. I couldn't have said it better myself. Shifty you seemed to have answered your own question.

Your question

this begs the question why Sony didn't just go with a dual Cell config and leave nVidia out of the picture

Your answer

I think the Cell demos were to show how powerful the Cell is ina way that can be communicated and be impressive. They could say 'we calculated Pi to 140,000 decimal places in one second' or something but that doesn't have oomph. By showing a software renderer, things devs know are complicated, running in realtime, they exhibit in a visual communication its capabilities

It not a direct answer, but you get the point.

 

[Shifty Geezer]

That's what I was saying Cell, although showing good visuals, surely can't be as good as nVidia's offering. By which we can conclude that it's worth in contributing to PS3 visuals must be somewhat marginalised. As such, the visuals demos of Cell were to provide a representation of it's capabilities that could be communicated in a way the audience would understand, not necessarily as a demo of how it would be used in PS3.

 

[j^aws]

CPU<=>GPU two-way comms has been expected since Hofstee's presentation last year...

Software rendering that is integrated with hardware acceleration...this waht we get with CELL<=>RSX

8)