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

[gosh]

no its just what i have gathered in layman terms looking at the past console history and the current next gen.

 

[jvd]

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

There isn't enough info about the rsx and cell to know how well it will do generating textures and feeding them to the rsx . It has the bandwidth to make it plausible however this bandwidht may be consumed for othre things .


Anyway whats really in question is how well either of these chips (x360 cpu and cell ) will do while having to deal with the strain of the normal game tasks they will be asked to perform

 

[gosh]

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

There isn't enough info about the rsx and cell to know how well it will do generating textures and feeding them to the rsx . It has the bandwidth to make it plausible however this bandwidht may be consumed for othre things .


Anyway whats really in question is how well either of these chips (x360 cpu and cell ) will do while having to deal with the strain of the normal game tasks they will be asked to perform

well im just feeding off the information about cell given around march and the rsx information given at E3. looking at 7800 and remarks by sony about RSX both are similiar except RSX will have either more pipes or a higher clock according to E3 information

 

[3roxor]

well im just feeding off the information about cell given around march and the rsx information given at E3. looking at 7800 and remarks by sony about RSX both are similiar except RSX will have either more pipes or a higher clock according to E3 information

It will have a higher clock for sure. The question is if there are other differences.

 

[gosh]

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

There isn't enough info about the rsx and cell to know how well it will do generating textures and feeding them to the rsx . It has the bandwidth to make it plausible however this bandwidht may be consumed for othre things .


Anyway whats really in question is how well either of these chips (x360 cpu and cell ) will do while having to deal with the strain of the normal game tasks they will be asked to perform

I dont think that will matter to a very important degree because its the responsibility of the GPU as to what and how the game is output on the screen. the CPU i think is at the backburner while what we see on screen is all done through the GPU, thats why we see the same system on a PC having a 9500 ATI giving 20 fps for doom 3 while the same system having a X800 will give more than 50 fps and a better looking game

 

[Shifty Geezer]

Laa-yosh has commented on his view that Cell isn't too effective for offline rendering as it's key strength, single-precision FP, isn't much used in offline rendering.

What DOES need single precision performance is realtime physics, graphics, audio, and other realtime work, where accuracy isn't as important as processing speed, and that's where Cell is very good.

Cell in it's current form looks better suited for realtime applications (games, content creation, codecs) than offline movie rendering to me.

 

[jvd]

No thats not why . Its clearly gpu limited . There are games in which you are clearly cpu limited and changing from a 9500pro to a x800xt pe wont increase your fps (though u may be able to increase your fsaa ) however changing your cpu to a faster one will increase your fps .

There are games that are gpu limited and thus going from a 3ghz cpu to a 3.8 ghz wont change your framerate . However changing cards would .

This is common in the pc market as games are made at diffrent times and components are upgraded at diffrent times .

 

[Nicked]

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

-The Xbox was more powerful for lots of reasons. RAM/CPU/GPU all played a part.
-CPU power is very important. If anything the EE of the PS2 should've taught you that. It had to pick up the slack that the GS simple couldn't handle, and PS2 holds up very well.
-We don't know enough about the RSX or Xenos to even say they are roughly equivalent in power.

I think most people would agree that a systems power comes as a sum of its parts, not just the GPU.

 

[jvd]

Cell in it's current form looks better suited for realtime applications (games, content creation, codecs) than offline movie rendering to me.

Yes its not great for dp . However it will still be capable of off line rendering . Hook it up to a huge pool of ram and let the program go . It may take alot of wind out of its sails in dp but it will still work for this .

In terms of real time rendering as i've siad there are alot of tasks the cpu has to do . So while a second cell chip may be stellar at doing procedural synthesis . While its under load doing game tasks , a.i , sound and other tihings it may not (Same goes with x360 cpu ) these chips will have a huge load placed onto them . So i highly doubt procedural synthesis will be hugely used . Perhaps for some things in limited quanity .

 

[Shifty Geezer]

jvd's comments just made me think, procedural texture synthesis doesn't need double precision, so though Cell might not be cost-effective for a render farm, it could provide a very useful support role if it could be integrated into the pipeline. Don't know if anyone would try to integrate it though.

 

[jvd]

As long as the rsx can write into the ram in cell while the cores are acessing that data also it shouldn't be a problem

For textures its even easier . They just need to be pushed down the flex io and be used by the rsx right then or stored in the gdr ram .

 

[EndR]

Will multiplatformgames not use Cell entirely due to high cost? Meaning that, maybe devs will use a same config as 360 thread wise (meaning that maybe some SPEs will be idle). The point of this is to make a game that is aprox identical on both hardwares and therefore have a chance of selling better. This seems maybe cheaper to do than invest time and money to really try to utilize the whole PS2 hardware, and this just to make a game "look better".

What im wondering is, will thirdparty-devs settle for a "360 config" (because PS3 seems to be able to match this) regarding their games? I see it making more sense from a business standpoint for thirdparty to make games that looks identical on both machines and have the possibility to hace access to both 360 and PS3 userbase than try to spend time and money on utilizing the whole Cell architecture..

(I hope I make sense here... )

 

[London Geezer]

Will multiplatformgames not use Cell entirely due to high cost? Meaning that, maybe devs will use a same config as 360 thread wise (meaning that maybe some SPEs will be idle). The point of this is to make a game that is aprox identical on both hardwares and therefore have a chance of selling better. This seems maybe cheaper to do than invest time and money to really try to utilize the whole PS2 hardware, and this just to make a game "look better".

What im wondering is, will thirdparty-devs settle for a "360 config" (because PS3 seems to be able to match this) regarding their games? I see it making more sense from a business standpoint for thirdparty to make games that looks identical on both machines and have the possibility to hace access to both 360 and PS3 userbase than try to spend time and money on utilizing the whole Cell architecture..

(I hope I make sense here... )

Of course multiplatform games will "use Cell". The games wouldn't work otherwise.
However, as has been said countless times already on here and other places, multiplatform games, at least at the beginning of the consoles' lives, will only use the easiest option, which is only one core on X360 and only the PPU core on PS3. Maybe marginally helped by the other cores on the CPUs.

As time goes on, tools and middlewares will improve to make sure that multiplatform games will use either console strength.

 

[Shifty Geezer]

In terms pf hardware usage, you'l get

first party > middleware > cross platform

I imagine. A developer embarking on a crosplatform and writing their own functions to utilise the diffrent hardwares would be taking on a lot of work for themselves. Over time though middleware add a degree of 'bloat' to a product, more effort can be invested into optimizing it for the different platforms. So for example, writing a fluid dynamic routing for a SPE and another XeCPU and integrating it's results into geometry creation would be a task that could be integrated into a single function in a middleware solution. The same code could be used to make use of the effect in the game without worrying about how it's implemented, while the middleware developer can tweak the different implementations on both platforms to make the most of different architectures.

It'll be interesting to see where in-house engine development goes this upcoming generation. All those resources could probably be spent more effectively in other areas of the product. How much will middleware come to dominate?

 

[3roxor]

Of course multiplatform games will "use Cell". The games wouldn't work otherwise.
However, as has been said countless times already on here and other places, multiplatform games, at least at the beginning of the consoles' lives, will only use the easiest option, which is only one core on X360 and only the PPU core on PS3. Maybe marginally helped by the other cores on the CPUs.

If that would be the case than practically those games won't use CELL. B.t.w the interview provides some hints that Sony does alot to prevent this from happening. [/b]

 

[Titanio]

A lot of middleware will be using cores beyond the PPE and one core on Xenon, so even early games will see some benefit. But for the most benefit, yeah, we'll have to wait a little while.

 

[blakjedi]

The good thing about xeCPU is that its symmetric... code that runs on one core can be fairly easily switched over to other cores if need be...... I wonder if the X360 OS can do this automatically or not?

 

[rendezvous]

Laa-yosh has commented on his view that Cell isn't too effective for offline rendering as it's key strength, single-precision FP, isn't much used in offline rendering.

What DOES need single precision performance is realtime physics, graphics, audio, and other realtime work, where accuracy isn't as important as processing speed, and that's where Cell is very good.

Cell in it's current form looks better suited for realtime applications (games, content creation, codecs) than offline movie rendering to me.

My bold

I don't know about that.
The books I have on offline rendering are using single precision floating point for the majority of their code.

 

[Shifty Geezer]

Single precision was common until recently. I know the major apps have included double precision of late. I dont know any more than that, or how the major CGI companies work with single/double precision. I'm going by Laa-Yosh's experience in these matters. His comments on the matter are buried somewhere in this forum...

 

[jvd]

We might see the xbox 360 become the defacto dev platform .

There are some reasons why this might happen

1) early start means more systems out

2) it will most likely be a bit less powerfull than the ps3

3) nintendo will be the odd duck out


reason 2 is interesting . Its alot easier to make your game around the slower less powerfull system and then move it to the faster system than going the other way around . Of course depending on how hard or dificult programing for the respective cpus are this could change