Technical investigation into PS4 and XB1 audio solutions *spawn

[Love_In_Rio]

That'll only be an issue for a few devs who developed for PS3 exclusively. Any cross-platform game that targeted PC can use the PC source.

Yes, that´s why i said first parties. I am sure Guerrrilla or Evolution studios engines struggles (if any) are for the cpu. You could see Guerrilla having the 6 cores full with jobs in their presentation. And is sure that those 6 cores full of jobs give them much less performance than 6 cell spus full of jobs (+1 PPE with two threads). They will have to change jobs to the gpu.

Anyway, from third parties claims till now I have read zero praises about the cpu and a lot of praises about the GPU ( even Guerrilla praised the GPU and nothing about the cpu ). For example today praised the GPU the Warframe developer:

http://www.vg247.com/2013/07/05/war...tech-dualshock-4-best-console-controller-yet/

That PS4 LibGCM/PSSL version must be kick ass.

But that's the issue of balancing resources anyway. If Sony put in a DSP instead of a couple of CUs, that'd be more sound potential but less opportunity for arbitrary compute. You can always add more, and the interest in console technology is what choices the designers made to address which compromises. It's all about picking a mark, balancing hardware complexity with efficiency with cost with ease of development.

i don´t know how big would be a decent sound dsp, are they very transistors heavy?. Last gen i think they spoiled it with the scaler. I hope they haven´t made the same this time with the dsp.

 

[Shifty Geezer]

DSPs can be very economical. As I say, though, you have to draw the line somewhere. Custom hardware is always more efficient for a given silicon budget, but it adds complexity and lack of versatility as that silicon only does one job. So as a system engineer, you have to decide how much silicon budget to spend on general purpose computing in the CPU, and how much to spend on fixed-function units like (de)compression blocks, audio processing, video en/decoding, raytracing units, hardware collision detectors, fixed lighting hardware, etc.

 

[Brimstone]

So how does SHAPE compare with the latest generation of Sound Blaster Z cards?

 

[Epicurus]

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?

That is nonsense. By audio raytracing, they refer to the technique of computing the impact of the game world's geometry and materials on the audio output given the player's location. Of course, you have to calculate such things on the actual game world representation which is custom and, thus, cannot be dealt with by fixed-function hardware.

This is nothing that audio DSPs, including the XB1s, are there for. DSPs can modify audio outputs based on parameters that are the result of raytracing calculations. But the calculation of these parameters can only be done by software.

 

[Shifty Geezer]

You misread. I didn't say "use the audio DSP's to raytrace audio."* Sony outlined all their hardware and what it brought to the table, such as the CUs being able to raytrace audio (I cite as an example of Sony's hardware exposure). They did not mention any DSP capabilities, such as the audio block being able to equalise, filter, and/or apply environmental audio transformations. Ergo the sensible conclusion is PS4 lacks a hardware DSP.

* Although your premise is wrong. Fixed function hardware can do anything you design it for. If someone wants to create a fixed-function audio raytracer, they can do so, and provide libs for the devs to supply game environment in a suitable form, although I doubt audio raytracing hardware would differ to any degree from raytracing hardware in general.

 

[Aeoniss]

How can a CPU from 2005/2006 be superior to a CPU now?

I will grant you that it's a laptop tablet CPU x 2 with some modifications by Sony\AMD, but is performance really that lackluster?

I've not heard anything criticizing the CPU, and I -have- heard praise by developers for it. I think Guerilla did, actually.

 

[Scott_Arm]

So how does SHAPE compare with the latest generation of Sound Blaster Z cards?

Here's the info on the audio block. SHAPE is just one of four blocks.

http://www.vgleaks.com/durango-sound-of-tomorrow/

Soundblaster really doesn't seem to give much info about their audio processors, at least that I could find. Their card can mix 128 3D audio voices. SHAPE handles 512 voices, but I don't know what the "3D" means in the Soundblaster Z case to make any kind of comparison.

 

[Love_In_Rio]

How can a CPU from 2005/2006 be superior to a CPU now?

I will grant you that it's a laptop tablet CPU x 2 with some modifications by Sony\AMD, but is performance really that lackluster?

I've not heard anything criticizing the CPU, and I -have- heard praise by developers for it. I think Guerilla did, actually.

212 Gflops vs 104.In theorical performance almost double.Far more difficult to program yes, but far more powerful too.And first parties had cell almost dominated.The subjet is thirds not.

 

[Kb-Smoker]

Digital Foundry:


http://www.eurogamer.net/articles/digitalfoundry-sony-details-final-ps4-spec

People forget this was running on a devkit that was not final. It didnt have the custom southbridge chip that will run the os.

 

[warb]

People forget this was running on a devkit that was not final. It didnt have the custom southbridge chip that will run the os.

That kind of fixed audio functionality would have been emulated, and not taking up a CPU core. As it was emulated on XBone dev hardware.

 

[Kb-Smoker]

That kind of fixed audio functionality would have been emulated, and not taking up a CPU core. As it was emulated on XBone dev hardware.

Emulated on what exactly? It was stated that the killzone demo didnt have access to the custom southbridge chip.


Here is what is in the dev kit:

SoC Based Devkit

Available January 2013
CPU: 8-core Jaguar
GPU: Liverpool GPU
RAM: unified 8 GB for devkit (4 GB for the retail console)
Subsystem: HDD, Network Controller, BD Drive, Bluetooth Controller, WLAN and HDMI (up to 1980×1080@3D)
Analog Outputs: Audio, Composite Video
Connection to Host: USB 3.0 (targeting over 200 MB/s),
ORBIS Dualshock
Dual Camera

 

[warb]

Why have developers manage CPU resources for audio if there's to be some fancy audio DSP in the console?

The XBone alpha kits apparently had another 8-core CPU to emulate the sound block.

 

[Kb-Smoker]

Why have developers manage CPU resources for audio if there's to be some fancy audio DSP in the console?

The XBone alpha kits apparently had another 8-core CPU to emulate the sound block.

Xbone dev kit had beast intel cpu and a very high end gpu to emulate the esram, move engines, and shape processor.

I never saw a report on an extra cpu just for sound block.

 

[warb]

Xbone dev kit had beast intel cpu and a very high end gpu to emulate the esram, move engines, and shape processor.

I never saw a report on an extra cpu just for sound block.

I've misremembered some posts here by (((interference))).

 

[Davros]

Their card can mix 128 3D audio voices

Each voice with up to 4 effects applied,
(iirc) that number goes up to 1024 if triggered by midi but then you can only use uncompressed audio (wav files)

 

[Cyan]

So how does SHAPE compare with the latest generation of Sound Blaster Z cards?

I don't know about Sound Blaster Z cards but bkilian said that it was an order of magnitude more powerful than the Sound Blaster XFi series, iirc.

The thing is that I thought an order of magnitude is 2x something, but 3dilettante explained to me that an order of magnitude is 10x something, which I learnt from him.

Live and learn, eh?

 

[Davros]

well billikan is basing that entirely on the fact that many years have passed since the xfi dsp was developed
and he is assuming newer = more powerfull
as of now we dont know the capabilities of the shape dsp yet apart from the number of simultaneous voices supported and compression/decompression
As for 512 voices how many of them can have effects applied to them simultaneously ? how many effects and what are they ? what is the instruction length for each voice ect
Hopefully either a programming guide or an api guide will surface and we will be able to find out
we dont know the number of transistors either

 

[warb]

It certainly sounds impressive put this way:

Yeah... considering the best soundblaster can mix 128 channels with 4 effects, I'm very impressed </sarcasm>
Put it this way, the performance described in that supposedly leaked doc could not be replicated on the 360, even if all 3 cores were used purely for the audio.

 

[expletive]

It certainly sounds impressive put this way:

The X360 CPU and 8 Jaguar cores both have theoretical performance of around 100 GFLOPS IIRC, is it fair to then assume that the audio capabilities of the XBO could not be replicated on 8 jaguar cores either?

 

[bkilian]

well billikan is basing that entirely on the fact that many years have passed since the xfi dsp was developed
and he is assuming newer = more powerfull

Nope, I'm basing it entirely on the fact that the xfi DSP is a 10000 MIPS part.
As far as SHAPE is concerned, it can mix thousands of channels, do a filter and volume change on thousands of channels, run an equalizer and compressor on hundreds of channels, and sample rate convert 512 channels per audio frame. And that's just one of the 5 components in the chip. Any one of the other 4 could easily equal what the xfi does.