So is there enough power to do great game audio on SHAPE without X1 having to hit the Jaguar CPU?
It seems on neogaf they've kind of started a "SHAPE is (mostly) only for Kinect" style narrative. Granted, they are pretty much always wrong...
Technical investigation into PS4 and XB1 audio solutions *spawn
- Thread starter Brad Grenz
- Start date
So is there enough power to do great game audio on SHAPE without X1 having to hit the Jaguar CPU?
It seems on neogaf they've kind of started a "SHAPE is (mostly) only for Kinect" style narrative. Granted, they are pretty much always wrong...
I think that's true and not true. SHAPE is one part of the audio block, and it is not used for Kinect at all. It is fixed function that should alleviate the CPU from having to do a lot (all) of the basic audio work. It sounds like the DSPs are not programmable, in terms of the devs having access. They're either used for Kinect or for audio compression/decompression. The scheduler and DMA should also lift some work from the CPU. So the good news is that in terms of audio, Kinect should come mostly free and a lot of the basic audio work the CPUs had to do last gen was moved onto a fixed-function audio block. We've heard about some games using a full core on the 360 for audio last gen. It would be interesting to know how much of that type of workload the fixed-function SHAPE block could cover.
SHAPE isn't for Kinect at all, in fact does Kinect even use any part of SHAPE?
There's parts (a lot) in the audio block that is for Kinect though.
That's the misconception it seems, people think that the whole audio block is called SHAPE, when only one section of it is actually the SHAPE unit.
There's parts (a lot) in the audio block that is for Kinect though.
That's the misconception it seems, people think that the whole audio block is called SHAPE, when only one section of it is actually the SHAPE unit.
Not really that as such.
For XB1 it would still need to use CPU and/or GPU to do wavetracing, EAX type audio.
PS4 will have to do the same and majority of what SHAPE does on CPU and/or GPU.
Didn't know for certain how much the PS4 audio chip could actually do, and if it could do anything other than the encoding and decoding that was mentioned, as Cerny's comment left it open that it may do a bit more.
interesting, i dont know anything about audio programming but how intensive is filtering and mixing? i ask because on vita the audio chip handles all of that i just imagined the same would go for ps4 but this seems like a step back in that regard. like if they ported vita games to ps4 would audio suffer if the game made heavy use on vita?
You should post a correction then. The fact that MS gives goofy names to pieces of hardware only adds to the confusion, people are generally just calling the audio hardware "SHAPE", but the name is irrelevant. The only reason they are saying that is because bkilian has said as much in this thread. He said (paraphrasing) that the audio block exists because they needed it for Kinect, they just happened to add some extra functionality to it since they were already designing and dedicating silicon for audio.
I'm sure he will correct me, which is great, my memory sucks.
Audio will be handled by a combination of Audio chips, CPU and CU on all systems. The balance will just be different. It will be interesting to see what the Xbox 1 will bring. On the PS3, the SPEs were giving the PS3 a huge advantage in audio processing in theory, but we haven't seen much more than exclusives make use of that, and we don't know if there's a bottleneck now for the PS4.
I'm from the days of the Amiga though and can definitely see the power of dedicated chips - but over time, also the power of flexible parts that can be allocated to different types of processing. So use-cases will have to prove which was the smarter decision.
It is clear though that Kinect integration required more dedicated silicon to perform various audio-processing tasks, and if Sony wants to replicate such functionality with their PS Eye, they will likely have a harder time doing so.
Personally, for games audio I think raycasting is the most important as it helps the most with creating realistic sound, and for this, CUs should be good. But I could be wrong, I'm not an audio expert.
I'm from the days of the Amiga though and can definitely see the power of dedicated chips - but over time, also the power of flexible parts that can be allocated to different types of processing. So use-cases will have to prove which was the smarter decision.
It is clear though that Kinect integration required more dedicated silicon to perform various audio-processing tasks, and if Sony wants to replicate such functionality with their PS Eye, they will likely have a harder time doing so.
Personally, for games audio I think raycasting is the most important as it helps the most with creating realistic sound, and for this, CUs should be good. But I could be wrong, I'm not an audio expert.
From his interview, Cerny's gamble is that future games will balance workloads between CPU and GPGPU, including audio work, ray casting, and things that are more efficient on a GPU. They claim to have modified the GPU specifically for that. The problem is that none of the launch games does yet. That's why I find it hilarious when we're supposed to deduce the console's power based on launch games. PS3 launch games didn't look very good in comparison to the ones that came later down the road, so did the PS2 games.
Yeah, it's their fault that they gave a design a reasonable name that was protected by NDA, and that it eventually leaked and people can't read. If people had read the actual leak, it's very clear that SHAPE (Scalable Hardware Audio Processor Engine) is just one of several pieces to the audio block, each of which is described in some detail. I do thing they should have dropped the first two letters in the acronym and called it APE.
ACP, AVP, ASP, SHAPE
There really isn't a whole lot that bkilian is saying that isn't described here: http://www.vgleaks.com/durango-sound-of-tomorrow/
If people would just read it instead of going by second, third, fourth etc hand information, then much of the discussion wouldn't be a problem.
ACP, AVP, ASP, SHAPE
There really isn't a whole lot that bkilian is saying that isn't described here: http://www.vgleaks.com/durango-sound-of-tomorrow/
If people would just read it instead of going by second, third, fourth etc hand information, then much of the discussion wouldn't be a problem.
It's a reasonable gamble of course since:
- AMD really wants compute to be a thing used by everyone.
- Nvidia has a business based on compute essentially. ( Titan is a reduced compute version of more compute heavy cards )
- Every GPU these days has at least some compute around to play with.
- XB1 also has compute.
So GPGPU isn't the easiest thing to program for but there are lot of people looking into it because it's ubiquitous and it can solve problems. Cerny is also betting that there will be middleware folks that may find compute a profitable niche. That would be more speculative but definitely a win for GPGPU if they did.
Well Resogun is a great showcase for compute and it is a launch title :smile:
Filtering and effects performed by CPU, or GPGPU. Not as fancy as SHAPE then.
curious why xbox 1 need to use cus for audio? for stuff like raycasting?
Depending on what you consider to be audio effects, SHAPE doesn't do them either.
I think what confuses some posters is that the word "processor" shouldn't be used to describe what is basically a group of fixed functions. Shape should be called Fixed Function Audio Group,while the other 3 are DSPs and they seem to have all the fexibility of a real DSP.
so kind of like the weakest link? do what you can audio wise on ps4s cpu then on the xbox 1 version you can move that to shape and free the cpu? ive always been told that because of shape, if you want to do equivalent audio on ps4 developers will have to use the 4cu. bkilians post about using less expensive/complex algorithms and doing those on the cpu would work fine for ps4 versions of games, makes sense to me but i've read others who say differently and that because the cpus are identical on both platforms that if they don't keep exact parity it means stuff like framerate loss.
ive been trying to find information on doing audio on gpus but i cant find anything, is it not a big thing or something?
also it was mentioned in the ps4 thread that even xbox 1 will have to use cpu and cus for audio in addition to shape just the ratio will be different than on ps4. i thought shape removed the need to use cpu or cu resources completely?
sorry if im asking too many questions.
I think what confuses them is their laziness in commenting on things without making an effort to understand them first. They mostly want a quick and easy answer that will help them build a case for their favorite console.
Depends what you mean by "great game audio" can shape decompress a hell of a lot of high quality simultaneous audio streams and place them in a game - yes (although limited to 48khz)
Can shape on its own do correctly environmentally modeled audio then no it will have to invoke the cpu (I dont see this as a problem, xb1 has enough cores)
I worry that some people think that because shape is very powerful its automatically capable of everything
like Cyan who said "shape would be able to do eax 50" when in fact shape cant do even eax 2.0 as it lacks some of the functions.
or if you like, to give you an analogy if you could overclock a geforce 2 by a factor of 50 it could do dx11 because it would have a hell of a lot of power
Similar threads
- Locked
