EA's Vision for the Future: Glenn Entis interview.

GameSpy: I hear you guys are the only company that has multiple PlayStation 3 dev kits.

Entis: We do have PlayStation 3 [kits]. I don't think I'm allowed to comment on any specifics. We clearly have PlayStation 3 dev equipment because we did a real-time PlayStation 3 demo at the Sony press conference. My team up in Vancouver has one. That was where that demo was developed in coordination with the Chicago-based Fight Night team. So we've had that for several months.

GameSpy: So having 5 times that floating point technology on the PlayStation 3 must be a...

Entis: Well, I like PlayStation 3 very much. We work with each of the companies in different ways. They all have different type machines and different market propositions. I'm looking forward to all of them.

GameSpy: Is PlayStation 3 more powerful than 360?

Entis: Well, I don't think any of us are really ready to say for sure. There's no final hardware specs on any of the systems [at the time of this interview].

Sony will have more processing power. There's no question about that. Then the arguments are about how easy it will be for people to get to and use it. The extra processing power will help. I think a lot of what we've been talking about in terms of dynamics is that physical simulations, fluid dynamics and to a certain extent AI. Some of the AI may actually end up using floating points, especially when you're using statistical AI and processing. Floating point is going to help all that.

The other place that I think floating points will help is less relevant to consumers. I think floating point will help with the process of building games. I think we'll see much more procedurally generated assets. So we'll see textures, noise patterns, and animation with some components that are procedurally generated. We'll see models procedurally generated in some cases, particularly when there are very large scale worlds with a lot of variety.

Every team is going to have a choice, "Do we actually build all that stuff one-by-one, or start abstracting out rules that can generate some part of those models on the fly?" All of that is going to start moving from how much raw data can we just take and play back as is?

Architecturally we're going to want to shift the problem of creating worlds from how much human labor and disk space we throw at it to how do we start condensing some of that and start moving on to the processors.

http://www.gamespy.com/articles/634/634928p1.html
 
Do you have any thoughts to provide on this topic? It would be nice to see your views.

The comments about the procedural animations and textures is neat. It is something that was hyped in the beginning years of the PS2 generation. I wonder if this will require more thoughton artists or better solutions from middleware to create these textures.
 
I do.

I wonder why MS thought they had "the" procedural synthesis solution this generation when cell offers so much more raw hp.

I wonder how 360 was designed for procedural synthesis (memexport?).
 
Ms has procedural synthesis . They will be quite capable of it . We don't know how either chip performs when doing this . So its hard to claim either way .


Anyway with the amount of compresion avalible on the x360 i would think we would still see a ton of textures the old way .
 
Well, I am sure the unified shaders on Xenos will be able to generate textures of some sort. What I would really like is real support for 3D textures with some type of compression scheme.
 
blakjedi said:
I wonder why MS thought they had "the" procedural synthesis solution this generation when cell offers so much more raw hp.

"the" should be "a".

They have 'a' solution that works within the larger framework of their goals. They went with a more traditional multicore/processor design instead of a stream processor (like CELL). MS has expressed the belief that while there will be an increase in floating point performance for procedural synthesis, textures, and so forth, they also believe that GP performance is still important. Neither MS or Sony has forsaken GP for FP or FP for GP. They just have different ideas of where the market is going and how best to get performance out of a given transistor budget.

One wrinkle I see, at least for procedural textures, is the jump from 256MB to 512MB. MS's CEO admitted this was a more recent design decision. 512MB does downplay the need for procedural textures, at least if bandwidth is not an issue.

Anyhow, MS's solution for procedural synthesis was the CPU cache locking to allow CPU<>GPU interaction (memexport probably does not hurt for other stuff). This will allow the CPU to dynamically generate meshes and feed them to the GPU.

Obviously CELL is well designed for this type of work, and according to DeanoC will probably outpace the XeCPU.

Of course that would be assuming that was if both CPUs were only doing procedural synthesis, which of course, they wont.

Each have design strengths, and considering CELL has 85M more transistors it should come out on top more often than not.
 
Acert93 said:
512MB does downplay the need for procedural textures, at least if bandwidth is not an issue.
Procedural synthesis isn't about saving RAM. I think that's just a happy byproduct. It's mostly about increasing realism by removing repeated patterns and objects. Varied trees, non-tiled grass and gravel surfaces, and small changes like this will elliminate a lot of current limitations in scene realism. Facial animations etc. are procedural synthesis of a different sort and again are needed for realism.

First titles will see little of this type of work I'm sure. Mastering the hardware and developing fast procedures will take time. But even if the consoles came with 10 GB of RAM there'd be need for procedural synthesis to get more realism.
 
Well, in lord of the rings, the massive armies were procedurally generated. Fire and smoke effects are usually procedural. Vegetation is another candidate.
 
I would think vegetation would be better if u used g.i

Anyway procedural synthesis will have its place . But i doubt either one will have enough power while running a game to use it all over the place
 
PC-Engine said:
Procedural synthesis has very few uses. Look at any CGI movie and tell me how much was procedurally generated.

Random, noise and detail textures will benefit. Vegetation will benefit. Developers will find ways to exploit it if they think it's worth their time.

I don't think we're at the stage yet, where we can say "look at how things are done on CG movies" as if to say we should do things that way. CG movies have a fair bit of stuff in them that won't be done in realtime for a long time, i don't see how this is different.
 
I don't think we're at the stage yet, where we can say "look at how things are done on CG movies" as if to say we should do things that way. CG movies have a fair bit of stuff in them that won't be done in realtime for a long time, i don't see how this is different.

But if you want it to look good then you're not going to depend on precedural stuff. Again it has its uses, but they're VERY limited. Why would CGI not use precedural synthesis if it make it look good in the majority of cases?

DemoCoder said:
Well, in lord of the rings, the massive armies were procedurally generated. Fire and smoke effects are usually procedural. Vegetation is another candidate.

What do you mean? A soldier has to be modeled. You cannot procedurally generate a soldier. Simply tossing in random elements does not equal procedural synthesis. There's very little procedural anything in Massivesoftware.

As to smoke, fire, and water, they're physics based not procedural.
 
gosh said:
looking at the data the PS3 raw numbers are good for "movies" and not for "games"

I don't understand what you mean . Any cpu chip is good for movies . Since its not in real time . They just take varying degrees of time to do whats needed .

Idealy you want a good balance between cpu power , ram space / speed and general i/o speed . You want it as balanced as possible . IF cpu power far out strips ram space / speed and i/o speed your just wasiting your time and money on the cpu . The same goes for any of those
 
jvd said:
gosh said:
looking at the data the PS3 raw numbers are good for "movies" and not for "games"

I don't understand what you mean . Any cpu chip is good for movies . Since its not in real time . They just take varying degrees of time to do whats needed .

Idealy you want a good balance between cpu power , ram space / speed and general i/o speed . You want it as balanced as possible . IF cpu power far out strips ram space / speed and i/o speed your just wasiting your time and money on the cpu . The same goes for any of those

I know but these raw numbers concerning Cell (not RSX) are good for generating CGI and Movies and not necessarily important in game output which is done by the way the GPU and CPU interact which is why the Xbox 360 and Playstation 3 will have similiar final output in the end
 
gosh said:
I know but these raw numbers concerning Cell (not RSX) are good for generating CGI and Movies and not necessarily important in game output which is done by the way the GPU and CPU interact which is why the Xbox 360 and Playstation 3 will have similiar final output in the end
Could you elaborate?
 
Playstation 2 CPU had more power than the Xbox cpu but why do Xbox games look/perform better? because of its GPU. we know that the GPU of RSX and Xenos have nearly the same output power so in the end the CPU power per say doesnt matter as long as the GPU handling it is nearly similiar in power
 
gosh said:
I know but these raw numbers concerning Cell (not RSX) are good for generating CGI and Movies and not necessarily important in game output which is done by the way the GPU and CPU interact which is why the Xbox 360 and Playstation 3 will have similiar final output in the end
Are you in the industry? Just wondering.
 
Back
Top