Watch Impress PS3 Technical article, from GDC (some new info)

Titanio said:
Ja, that would definitely be one ever-present requirement, although it wouldn't surprise me totally if that was something RSX might handle (?). I think more generally they're serious about have PS3 as an "always/on connected" server and media player, even if someone is playing a game at the same time.

HD scaling is pretty computationally intensive, i dont think its something theyd want to spend RSX cycles on. The comments on 'why does the OS need a whole SPE?' just got me thinking, using one SPE for the OS AND scaling would be a good use for it, if they MUST dedicate one. The OS on the 360 takes less than 5% of one core so i would assume a dedicated SPE would have lots of free time.
 
expletive said:
HD scaling is pretty computationally intensive, i dont think its something theyd want to spend RSX cycles on.

They may have a dedicated fixed function scaler, you never know. I think Cell is a good candidate for this, no doubt, but we just can't be too sure yet what does what.

expletive said:
The comments on 'why does the OS need a whole SPE?' just got me thinking, using one SPE for the OS AND scaling would be a good use for it, if they MUST dedicate one. The OS on the 360 takes less than 5% of one core so i would assume a dedicated SPE would have lots of free time.

You have to accomodate the worst case. In PS3's case, that could well be decoding a 1080p movie stream, scaling the framebuffer, and playing the game at the same time - not trivial, very suddenly you'd be needing the guts of a SPE, I'd bet. You might consider the SPE 'wasted' if you're not doing all those things, but they have to be accomodated if the system is to be capable of that.

Plus it gives them perhaps more room for future software upgrades.
 
Titanio said:
You have to accomodate the worst case. In PS3's case, that could well be decoding a 1080p movie stream, scaling the framebuffer, and playing the game at the same time - not trivial, very suddenly you'd be needing the guts of a SPE, I'd bet. You might consider the SPE 'wasted' if you're not doing all those things, but they have to be accomodated if the system is to be capable of that.

Plus it gives them perhaps more room for future software upgrades.

Any word on where the PS3 is decoding/steering in-game multichannel audio?
 
That'd be on Cell, I'm sure, something of application concern probably. There was talk before of "RSX audio", but I don't think that RSX was referring to the GPU (there's some audio middleware called RSX, AFAIK).
 
While I believe one SPE will be dedicated to the OS/Security/Etc., I can't see that SPE being reserved for just those things. I'm sure the developers will be able to utilize all 7 SPEs, as well as the PPE, for gaming application. The CELL's SPEs stream through data being delegated to them. Just because one of the SPEs is handling OS/Security/Etc. doesn't mean that when that SPE is idle it cannot be used for anything game related.
 
Titanio said:
That'd be on Cell, I'm sure, something of application concern probably. There was talk before of "RSX audio", but I don't think that RSX was referring to the GPU (there's some audio middleware called RSX, AFAIK).

DeanA was quite assertive that it had nothing to do with the Miles Sound System RSX 3D Audio.

The diagram we saw from GDC showed that RSX has to at least deal with the 7.1 audio.
 
Titanio said:
That'd be on Cell, I'm sure, something of application concern probably. There was talk before of "RSX audio", but I don't think that RSX was referring to the GPU (there's some audio middleware called RSX, AFAIK).

Then there's another thing that would be a good candidate for an "OS SPE". I agree with your point on planning for the 'worst case scenario' too. Game developers need to plan for the fact that people may be using the other multimedia functions of the console to their fullest when a game is running.
 
Last edited by a moderator:
ROG27 said:
While I believe one SPE will be dedicated to the OS/Security/Etc., I can't see that SPE being reserved for just those things. I'm sure the developers will be able to utilize all 7 SPEs, as well as the PPE, for gaming application. The CELL's SPEs stream through data being delegated to them. Just because one of the SPEs is handling OS/Security/Etc. doesn't mean that when that SPE is idle it cannot be used for anything game related.

The PPE is certainly useable, that's not in question. I'd venture to guess the OS only uses a very very small percentage of its time.

The thing is, if a certain amount of work has to be accomodated at any point in time during gameplay, the amount of resource required to power that must be reserved at all times, even if it's not being used. Certainly, the developer cannot rely on it, and thus even if it could be used during 'idle time' I'd say most would probably ignore it without guarantees about its availability.

expletive said:
Then there's another thing that would be a good candidate for an "OS SPE".

While it's possible the OS could provide some "bog-standard" audio processing services to the game, I think many apps would require more than it might be able to offer from a significantly shared SPE - hence why I think how audio processing is done would be of the application's concern. We know, for example, that DeanoC has said they're dedicating an entire SPU to audio in Heavenly Sword, and they don't seem to be alone in such intensive CPU use for audio.
 
Last edited by a moderator:
Mmmkay said:
DeanA was quite assertive that it had nothing to do with the Miles Sound System RSX 3D Audio.

The diagram we saw from GDC showed that RSX has to at least deal with the 7.1 audio.

Could that make RSX some sort of a 'southbridge' which controls peripheral i/o, audio, and a GPU? Certainly NVidia has plenty of experience in regard to core logic chipsets to design such a chip...
 
Mmmkay said:
DeanA was quite assertive that it had nothing to do with the Miles Sound System RSX 3D Audio.

The diagram we saw from GDC showed that RSX has to at least deal with the 7.1 audio.

Hmm, I hadn't seen that (and it was in response to me - eep! :oops: ). That's kind of interesting. Where's the slide, also?
 
expletive said:
Could that make RSX some sort of a 'southbridge' which controls peripheral i/o, audio, and a GPU? Certainly NVidia has plenty of experience in regard to core logic chipsets to design such a chip...

Not sure, but I don't think so. The old dev kit diagrams showed a southbridge:
http://pc.watch.impress.co.jp/docs/2005/0722/kaigai02l.gif

As did the diagram which was presented at GDC:
http://www.watch.impress.co.jp/game/docs/20060329/3dps303.jpg <-- This slide Titanio
 
DeanA's comment aside, does where the audio is outputted from tell us where it is processed? Is that referring to HDMI audio, is other audio e.g. the digital-audio out, outputted elsewhere?
 
Mmmkay said:
Not sure, but I don't think so. The old dev kit diagrams showed a southbridge:
http://pc.watch.impress.co.jp/docs/2005/0722/kaigai02l.gif

As did the diagram which was presented at GDC:
http://www.watch.impress.co.jp/game/docs/20060329/3dps303.jpg <-- This slide Titanio

Interesting...

That would possibly explain some disparity on transistor counts and pipelines when compared to stock PC parts...

EDIT: also, based on the 2nd diagram it seems as if the scaling happens 'to the right' of RSX or in RSX itself as Titanio suggested, which would leave the cell out of the scaling equation.
 
Last edited by a moderator:
Isn't it obvious? If the PS3 is going to be a location free server for the PSP it's going to need to re-encode video on the fly, while people are potentially playing games on that same PS3. They can't saddle that on the PPE cause it would probably be too hard to manage the performance drops in-game, by reserving an entire SPE they make it easy for dev's to target the full power of the system without having to worry about how it will perform in various situations.
 
Last edited by a moderator:
scooby_dooby said:
Isn't it obvious? If the PS3 is going to be a location free server for the PSP it's going to need to re-encode video on the fly, while people are potentially playing games on that same PS3. They can't saddle that on the PPE cause it would probably be too hard to manage the performance drops in-game, by reserving an entire SPE they make it easy for dev's to target the full power of the system without having to worry about how it will perform in various situations.

I wonder if they need to reserve portions of inter-chip bandwidth as well for such activities?
 
Last edited by a moderator:
*possible stupid question*

The ps3 has 256MB of XDR@3.2ghz.8x the speed,1/8th the quantity of pc

My pc has 2GB@DDR400mhz.(not really:rolleyes: )

Are they comparable?Does it work like that?

I read this on another site..........
 
No it doesn't work like that at all. The XDR is better performing so as long as the RAM required for 'x' task is under 256MB (for the purposes of your example), the XDR will have a clear advantage over the PC with the slower 2GB of RAM. However, the amount of RAM puts a very real 'cap' on what the system itself is able to work on - of course there's virtual cache out on the hard drive - and if the apps running require more than the hypothetical 256MB of RAM, the advantage is clearly with the greater, albeit slower, RAM in the PC. More RAM is better not as an alternative to faster RAM, but because the amount is the main determiner of what is even possible on the system. The speed of the RAM more effects how well those given applications run. So it's really case by case as to which you would prefer, but for a gaming platform clearly the system with greater amounts of RAM will have more doors open to it.
 
so all we really know about RSX is that it can do 24 texture look-ups per clock, which I assume means it has 24 texture units.

we don't know for certain how many 'pixel shading pipes' or 'vertex shaders' it has, or how many ROPs it has.

so far, the best guesses are, 24 pixel shader pipes, and 8 ROPs.

what about this possibility: 12 pixel shader pipes, with 2 textures each, making 24 textures, 8 ROPs, and 6 vertex shaders. just a wild guess. probably wrong. but something has to give, given the 128-bit bit memory interface.
 
xbdestroya said:
No it doesn't work like that at all. The XDR is better performing so as long as the RAM required for 'x' task is under 256MB (for the purposes of your example), the XDR will have a clear advantage over the PC with the slower 2GB of RAM. However, the amount of RAM puts a very real 'cap' on what the system itself is able to work on - of course there's virtual cache out on the hard drive - and if the apps running require more than the hypothetical 256MB of RAM, the advantage is clearly with the greater, albeit slower, RAM in the PC. More RAM is better not as an alternative to faster RAM, but because the amount is the main determiner of what is even possible on the system. The speed of the RAM more effects how well those given applications run. So it's really case by case as to which you would prefer, but for a gaming platform clearly the system with greater amounts of RAM will have more doors open to it.

Thanks for that!:p
 
Megadrive1988 said:
what about this possibility: 12 pixel shader pipes, with 2 textures each, making 24 textures, 8 ROPs, and 6 vertex shaders. just a wild guess. probably wrong. but something has to give, given the 128-bit bit memory interface.

Nothing has to 'give', but a reduction in ROP-count seems reasonable. But two texture units per PS seems a little crazy to me, I mean the most obvious answer is staring us in the face ;)

expletive said:
I wonder if they need to reserve portions of inter-chip bandwidth as well for such activities?

Probably a certain amount of all bandwidth is reserved, and this is achievable according to IBM.
 
Back
Top