Technical investigation into PS4 and XB1 audio solutions *spawn

[ERP]

Then you also have custom audio part, which could free up CPU resources.

I really don't think that the difference will be as large as many seem to be assuming.

Does the dedicated audio silicon on XB1 free up CPU resources, giving it an advantage in CPU limited scenarios?

 

[Shifty Geezer]

You've lost me there. Since PCM samples we've been able to arbitrarily play back samples at any frequency. It's the basis of sample-based music going back to PC wavetable cards and E-MU and Fairlight before that. I can stick a sample on a key and pitch bend it in perfectly adequate quality, and could do that on the Amiga (MED tracker with pitch shift instructions). I can't see anything stopping a racer from the past 10 years getting the relative velocity between a car making a sound and the player and pitching the sample accordingly.

 

[Arwin]

Old school, that method would slow down the sample to get the pitch lower. It would lengthen the sample, and the information per second is less. Modern hardware can change the pitch without slowing down, preserving information density and duration. A cool example I heard was changing George Michael to Kylie Minogue's pitch and vice versa, showing that their method and quality of singing was near indistinguishable.

Anyway, maybe that's the difference in quality bklian is referring to?

I personally think the doppler effect could easily and by far the most efficiently with the least amount of quality change be accomodated by the car-sound simulation engine, if that can take into account environmental factors by default. Say what you will about Gran Turismo's sound but if you listen to the cars driving around on the track inreplays or when you have the pre race screen showing the track with cars racing around, most doppler and environmental effects are handled quite well already. The current synth is mostly held back by really weak low frequency effects. Particularly the impact of the exhaust, gas explosion and valve movement is too underrepresented, as is simple variation in volume from car to car.

 

[bkilian]

Old school, that method would slow down the sample to get the pitch lower. It would lengthen the sample, and the information per second is less. Modern hardware can change the pitch without slowing down, preserving information density and duration. A cool example I heard was changing George Michael to Kylie Minogue's pitch and vice versa, showing that their method and quality of singing was near indistinguishable.

Anyway, maybe that's the difference in quality bklian is referring to?

I personally think the doppler effect could easily and by far the most efficiently with the least amount of quality change be accomodated by the car-sound simulation engine, if that can take into account environmental factors by default. Say what you will about Gran Turismo's sound but if you listen to the cars driving around on the track inreplays or when you have the pre race screen showing the track with cars racing around, most doppler and environmental effects are handled quite well already.

I thought it was Jason Donovan and Kylie Minogue, but yes. Having written my own MOD and S3M tracker/Player, I have some experience with resampling. MOD (from the Amiga) and S3M (From the PC) both use nearest neighbour resampling. Since they're playing with electronic sounding music in general, it doesn't hurt too much, and it's very cheap. It introduces some horrible artifacts though, and doesn't work for samples too far away from their original source frequency (thus why piano sounds on MODS always sound crap unless they have multiple samples at different frequencies, like klim.mod)
Most game audio engines use linear interpolation resampling, in which they take the average of the two source samples they come in between. It can be a little more sophisticated than that, because you know basically where in the continuum of the two samples you've landed, so you can use the classic "pick a point on the line between A and B that corresponds to your X position".

Unfortunately, linear resampling causes aliasing of the signal, and introduces harmonics above nyquist. The resampler in the One is much higher quality, and would be very expensive if done in software.

 

[Shifty Geezer]

The quality issue with slowing down samples isn't terrible. It's still used to the this day in digital music with professional wave-based software instruments, and it's still used on lots of game samples. And pitch shifting without changing speed tends to come with a lot more artefacts. Being able to repitch a sound may lead to better quality, but I don't see that as an excuse not to implement Doppler in simpler ways.

 

[Davros]

And pitch shifting without changing speed tends to come with a lot more artefacts

Ive not noticed that, for a laugh Ive just played quake 3 pitch shifted (both up and down) and avp colonial marines

ps: quake live supports doppler

 

[Love_In_Rio]

Does the dedicated audio silicon on XB1 free up CPU resources, giving it an advantage in CPU limited scenarios?

Hasnt Sony also an audio chip?.Seems you imply not.

 

[Aeoniss]

Hasnt Sony also an audio chip?.Seems you imply not.

PS4 has an Audio DSP but there's not a whole lot of info on it. At the very least, it seems that the SHAPE block is significantly more powerful.

On the other hand, it's likely that Kinect (which is hooked up to the system 24/7) will be using it and has significant portions of its resources 'reserved' for itself. In the same way that, likely, much of the OS memory reservations are for Kinect and its functionality.

So when it comes to "game time" what will SHAPE be able to do in non-Kinect scenarios (excluding OS stuff) scenarios?

Zero idea.

 

[Love_In_Rio]

PS4 has an Audio DSP but there's not a whole lot of info on it. At the very least, it seems that the SHAPE block is significantly more powerful.

On the other hand, it's likely that Kinect (which is hooked up to the system 24/7) will be using it and has significant portions of its resources 'reserved' for itself. In the same way that, likely, much of the OS memory reservations are for Kinect and its functionality.

So when it comes to "game time" what will SHAPE be able to do in non-Kinect scenarios (excluding OS stuff) scenarios?

Zero idea.

Well, if the chip wont free cpu resources or its a very simple mp3 decoder or its not finally inside.There was here a crazy rumor (a friend of a friend) saying helper blocks didnt make the cut (audio , zlib decompressor and scaler).

 

[Maggie]

PS4 has an Audio DSP but there's not a whole lot of info on it. At the very least, it seems that the SHAPE block is significantly more powerful.

On the other hand, it's likely that Kinect (which is hooked up to the system 24/7) will be using it and has significant portions of its resources 'reserved' for itself. In the same way that, likely, much of the OS memory reservations are for Kinect and its functionality.

So when it comes to "game time" what will SHAPE be able to do in non-Kinect scenarios (excluding OS stuff) scenarios?

Zero idea.

Why? Because it has a nice name and Sony's chip not? Or because it could be one of the last possibilities where Microsoft's "secret sauce" is hidden? Do we have any facts about those chips yet? I think not, that's why such statements are a bit silly.

 

[Shifty Geezer]

Why? Because it has a nice name and Sony's chip not? Or because it could be one of the last possibilities where Microsoft's "secret sauce" is hidden? Do we have any facts about those chips yet? I think not, that's why such statements are a bit silly.

Lose the sarcastic comments. This forum's becoming a juvenile bitch-fest at the moment.

The reason to believe XB1 has a significantly more potent audio processor is because the guy who helped design it has told us it was a priority component for MS and it was designed with particular sophisticated features. At this point all we know of PS4's audio is it has an audio decoding block that can blend multiple samples. Nothing has been said of PS4's audio aspect doing audio processing, so, although it's an assumption, it's an assumption based on provided information, and a stronger assumption than PS4's audio block being SHAPE's equal based on nothing whatsoever.

 

[Maggie]

So the assumption is that because we know more about the Xbox chip than we know about the PS4 chip, it has to be more powerful? I don't think that's a safe assumption at all.

 

[Shifty Geezer]

No, the assumption is that putting advanced audio processors into boxes is rare and there's no evidence to support the notion that that's what Sony did, but it's known that's in XB1. It's not a 'safe' assumption, but the probabilities favour SHAPE > PS4audio.

You asked, "do we have any facts about those chips yet?" Go search this forum for posts by bkilian, an audio engineer who helped design SHAPE, who gives us reasonable detail on the capabilities of the processor. It's a significant DSP.

 

[Love_In_Rio]

So the assumption is that because we know more about the Xbox chip than we know about the PS4 chip, it has to be more powerful? I don't think that's a safe assumption at all.

Guerrilla Killzone SF presentation having almost a core doing sound processing tells a lot.Shape doesnt need cpu processing at all.One reason more to have a 1,8-2 GHz cpu instead of the paltry 1,6 (remember the same that the weakest Kabini flavour!).

 

[expletive]

No, the assumption is that putting advanced audio processors into boxes is rare and there's no evidence to support the notion that that's what Sony did, but it's known that's in XB1. It's not a 'safe' assumption, but the probabilities favour SHAPE > PS4audio.

You asked, "do we have any facts about those chips yet?" Go search this forum for posts by bkilian, an audio engineer who helped design SHAPE, who gives us reasonable detail on the capabilities of the processor. It's a significant DSP.

And this.

 

[Maggie]

No, the assumption is that putting advanced audio processors into boxes is rare and there's no evidence to support the notion that that's what Sony did, but it's known that's in XB1. It's not a 'safe' assumption, but the probabilities favour SHAPE > PS4audio.

They did not do this last gen, that's true. My assumption is, that, because both companies are using the same CPU, both companies realised, that the CPU wasn't perfectly suitable for audio processing, so both decided to use a dedicated chip (which seems to be true). But why use a dedicated audio chip when it's barely more powerful than a CPU core? And let's not forget that Sony is also an audio company, and they do not only produce audio hardware for professionals and consumers, but they also have artists who make music, movies and video games. I'm not sure if there is even any other company which is in such a unique position.

You asked, "do we have any facts about those chips yet?" Go search this forum for posts by bkilian, an audio engineer who helped design SHAPE, who gives us reasonable detail on the capabilities of the processor. It's a significant DSP.

Sure, but that still tells us not much about the PS4 audio processor, which in my opinion makes it hard to compare them.

Guerrilla Killzone SF presentation having almost a core doing sound processing tells a lot.Shape doesnt need cpu processing at all.One reason more to have a 1,8-2 GHz cpu instead of the paltry 1,6 (remember the same that the weakest Kabini flavour!).

I don't think that's final hardware. We know the PS4 also has an audio processor. So why have one built-in when you need a whole CPU core to help? That doesn't make much sense, sorry.

 

[TheWretched]

To be fair, Sony has a heritage of audio equipment. So, it's not unusual to assume, they could repurpose some of their own hardware here.

 

[Love_In_Rio]

To be fair, Sony has a heritage of audio equipment. So, it's not unusual to assume, they could repurpose some of their own hardware here.

With Yamaha chips...As a curosity Kutaragi designed one of the videogames history best sound dsps, the SNES one!.

 

[Shifty Geezer]

I don't think that's final hardware. We know the PS4 also has an audio processor. So why have one built-in when you need a whole CPU core to help? That doesn't make much sense, sorry.

We had Sony talking about using CU's for raytracing audio. Why wouldn't they talk about the potential of their audio DSP if it's capable of more than just what they've said about it?

As for Sony being an audio company, that's not related to their gaming sector. Sony's split into a lot of division with little crossover (that's changing) - eg. Just because there's a scaling chip used in Sony TVs, didn't mean they'd add one to their PS3 even when MS did. And as for both companies coming to the same conclusion, what's wrong with thinking of just using the CPU? As I've discussed elsewhere, if half the tasks that were CPU hogs have moved to GPGPU this gen, the workloads for the CPU cores to undertake are getting fewer. I don't see much wrong with allocating a core for audio when designing the system. An audio DSP is of course better, but you have to draw the line somewhere.

 

[bkilian]

The quality issue with slowing down samples isn't terrible. It's still used to the this day in digital music with professional wave-based software instruments, and it's still used on lots of game samples. And pitch shifting without changing speed tends to come with a lot more artefacts. Being able to repitch a sound may lead to better quality, but I don't see that as an excuse not to implement Doppler in simpler ways.

Professional wave based software synths tend to have a bank of samples per instrument. For piano, they sample at every octave, and sometimes even more often, especially for the higher frequencies. If you take a C3, and resample it to C4, you've halved the play time of the original note, including the decay, which in a real piano is normally about the same length no matter what note you're playing.