Cell + sound processing questions

[Andy]

I was reading that the cell processor in PS3 will also be the audio processor, I was wondering what the pros and cons would be for taking such an approach, which leads me to several questions:

1) Have any of the other consoles made from Sony or any other console manufacturer for that matter used this method before? Will XBox 360 (would say Revolution as well, but not enough known at this stage) be doing this as well, or will it have it's own dedicated audio processor?

2) In all likelyhood how much of a performance hit will cell take now that it has to do audio as well as things like AI, physics etc. ?

3) What benefits does this having allowing the CPU to process sound rather than a dedicated audio processor?

Thanks for your time guys. All the best.

 

[London Geezer]

I was reading that the cell processor in PS3 will also be the audio processor, I was wondering what the pros and cons would be for taking such an approach, which leads me to several questions:

1) Have any of the other consoles made from Sony or any other console manufacturer for that matter used this method before? Will XBox 360 (would say Revolution as well, but not enough known at this stage) be doing this as well, or will it have it's own dedicated audio processor?

2) In all likelyhood how much of a performance hit will cell take now that it has to do audio as well as things like AI, physics etc. ?

3) What benefits does this having allowing the CPU to process sound rather than a dedicated audio processor?

Thanks for your time guys. All the best.

1) X360 does audio in the XCPU too.
2) very very very very small for what they'll have to do, which is 5.1 DD or DTS. Obviously if one were to have a 284.79 DTS soundtrack with each channel with a 18Mbps track, that might be a bit much.
3) more flexibility, cheaper as there is no sound chip to pay for and attach to the platform. More flexibility. And it's cheaper. Oh and flexible.

 

[_xxx_]

Sound processing won't take but the fraction of processing power available in PS3, hence the lack of a dedicated sound chip. Not really needed.

I also doubt that XBox360 will have a separate sound chip.

You can see the disadvantage of this approach on your average PC if you're using an integrated codec of your mobo chipset (other than NF3 APU), it swallows quite a few processor cycles on your average PC.

 

[London Geezer]

MS has already confirmed that the X360 will not have a separate audio chip.

Audio chips are sooo 90's.

 

[rabidrabbit]

Has there been any talk of the PS2 chip [EE+GS] being inside PS3, like the PSOne chip was inside PS2 for I/O, and I think you could use the PSOne chip for PS2 sound processing too, am I wrong?

Wouldn't the [EE+GS] be more than sufficient for full sound processing in PS3, while also doing those I/O fundctions?

 

[Nite_Hawk]

I was reading that the cell processor in PS3 will also be the audio processor, I was wondering what the pros and cons would be for taking such an approach, which leads me to several questions:

1) Have any of the other consoles made from Sony or any other console manufacturer for that matter used this method before? Will XBox 360 (would say Revolution as well, but not enough known at this stage) be doing this as well, or will it have it's own dedicated audio processor?

It probably depends on what exactly you mean by this. The xbox had a custom dsp (the soundstorm basically) for the xbox. I'm not entirely sure how programmable it is, though it was good enough to do the realtime dolby digital encoding. The PS2 also has a custom sound chip, and again I'm not entirely sure how programmable it is. It seems like it may be less so than nVidia's, but I'm not totally sure.

Plenty of PCs basically do thier sound processing using the host CPU. They tend to use a lot of system resources, but then most PC processors aren't particularly good at this sort of thing.

2) In all likelyhood how much of a performance hit will cell take now that it has to do audio as well as things like AI, physics etc. ?

It reallly depends on what you want to do. If all you are doing is playing back stereo pcm data it's probably going to be pretty minimal. If you want to do realtime 3D sound rendering (raycasting techniques) with 8 channels of sound and encode it into one of the various dts formats it will end up taking more. A lot of it will come down to how well implemented in the algorithms are. It'll be a balancing act, though luckily cell should be fairly good at this sort of thing (much much better than typical PC cpus).

3) What benefits does this having allowing the CPU to process sound rather than a dedicated audio processor?

Flexibility. If your game only needs basic stereo sound the cell processor can be doing other things where a dedicated audio processor would be sitting idle. On the other hand if you need very advanced sound and don't mind giving up a bit of processing power to get it you can do that too. Additionally, as new audio formats come out it will be easier to support them (though if your dedicated audio processor is reasonably flexible this might not be an issue either).

I think traditionally it has made a lot more sense to use dedicated audio processors when the host cpu isn't efficient at dealing with the kind of processing audio requires. Cell happens to be pretty good at it though, so there isn't as much need for dedicated hardware.

Nite_Hawk

 

[London Geezer]

Has there been any talk of the PS2 chip [EE+GS] being inside PS3, like the PSOne chip was inside PS2 for I/O, and I think you could use the PSOne chip for PS2 sound processing too, am I wrong?

Wouldn't the [EE+GS] be more than sufficient for full sound processing in PS3, while also doing those I/O fundctions?

EE+GS all for sound is like driving a Ferrari to go get milk across the road.

 

[Carl B]

EE+GS all for sound is like driving a Ferrari to go get milk across the road.

But sometimes that milk really needs to get there!

 

[London Geezer]

EE+GS all for sound is like driving a Ferrari to go get milk across the road.

But sometimes that milk really needs to get there!

Ok.

Like using an 18 wheeler to get the half pint of milk down the road.

 

[Carl B]

EE+GS all for sound is like driving a Ferrari to go get milk across the road.

But sometimes that milk really needs to get there!

Ok.

Like using an 18 wheeler to get the half pint of milk down the road.

But, sometimes....

LOL, anyway yeah sound will just be slaved away to one of the SPE's - as far as resource overhead the load will be immaterial.

 

[London Geezer]

EE+GS all for sound is like driving a Ferrari to go get milk across the road.

But sometimes that milk really needs to get there!

Ok.

Like using an 18 wheeler to get the half pint of milk down the road.

But, sometimes....

LOL, anyway yeah sound will just be slaved away to one of the SPE's - as far as resource overhead the load will be immaterial.

I mean don't get me wrong, i'm sure that if a programmer were to really work on an amazing super advanced sound engine, the CPU could be used properly. But he'd need to do some pretty wild stuff.
For everyday use, the load is really minimal.

I mean, if realtime in-game DTS encoding on PS2 (GTA3 to San Andreas, SSX Tricky and later and other games) took about 4% of the EE's VUs (according to archie), then i can only imagine how little it will afftect PS3 and X360.

 

[Carl B]

Oh I hear ya London - I never disagreed in the first place.

It's just that Ferrari analogy forced me to post in this thread.

 

[London Geezer]

Oh I hear ya London - I never disagreed in the first place.

It's just that Ferrari analogy forced me to post in this thread.

Well today that EE+GS is more Fiat Uno than Ferrari.... but still

 

[Nightz]

The N64 used the CPU for audio first and that caused problems.
I dont think you can ever have enough processing power no matter how trivial the task.

 

[London Geezer]

The N64 used the CPU for audio first and that caused problems.
I dont think you can ever have enough processing power no matter how trivial the task.

Well the N64 wasn't a PS3 or X360. I'm not sure if "looking at history" is the best course of action, and it seems people tend to use that methodology a lot around here.

The point is, sound on the next gen CPUs will be a trivial task, considering the kind of performance expencted. And it's cheaper not to include a separate sound chip which might or might not make things better in the end.