Bonaire, in the form of a 7790 equivalent, is more like "GCN 1.1" (or, as Anandtech put it, a slightly improved take on GCN 1.0)
Well, in fairness, they also had very little info to go on as AMD didn't tell them everything there was to know. For example, the 7790 is confirmed to have support for a more advanced form of Partial Resident Texture or Tier 2 PRT, which apparently none of the other GCN cards till this new lineup of cards possessed. And according to a Dave Baumann post I read a while back, he makes it sound like a variety of not so insignificant changes could be made without their necessarily being announced since GCN was always expected to be a constantly evolving and changing architecture.
The fact that this true audio stuff was apparently hidden in Bonaire this entire time is potentially another good example of just how much can be done without getting very much fan fair in the here and now when the card was announced and released.
Yes, most of it is reserved for Kinect processing, but you are missing some crucial context to that post. That post was in response to another user that directly referenced a previous post by bkilian where he told people just how powerful the
entire Xbox One audio block was.
Here is the post in question where he tells people how powerful the entire xbox one audio block is.
http://beyond3d.com/showpost.php?p=1787287&postcount=572
Sorry, I should have said "consumer sound cards". The X1 is not a professional mixing station, it's a game console.
And the mix buffers have 128 physical buffers, but can be used with over 4000 virtual buffers per audio frame. Think of them as registers that can hold an entire audio frame. The 21369 has 32, much smaller ones. The SRC can process 512 channels per audio frame, and the XMA decoder can decode 512 channels per audio frame.
The clock speed of the audio block is twice that of a 21369, and the fixed function blocks were calculated, per the hotchips presentation, to be 18 GOPS equivalent. The 21369 is 2.4 GFLOPS. If you assume the scalar tensilica cores are about the same power per clock of a 21369, and use the 15.4 GFLOPS value for the two vector cores, you're talking 23 21369s equivalent for the whole audio block. How much did that 12 core sound card cost again? I found an 8 core one for something like $1500. Let's change my statement to "the Xbox one audio block is far more powerful than any sound card you can buy for less than or equal to the price of an entire Xbox one."
I believe you heavily underestimate the power of the X1 audio block.
All that post you referenced is saying is that, compared to all the power he just described here in the above post, yes, developers truly do only have a small part of it. The entirety of the SHAPE audio block is
included in that small part.
SHAPE is not packing anywhere the kind of power that those 4 DSP cores, especially the two Vector DSPs, are packing. So even with devs having full access to SHAPE, that still only constitutes but a small portion of the full power inside the entire Xbox One audio block. You seem to be taking that post in isolation while overlooking the very important context that goes with it, I'm afraid.
Bkilian made the post you reference in direct response to this post, where onq seemed to believe that all the power in the xbox one audio block just described by bkilian would somehow be fully available to developers, which it won’t be.
http://beyond3d.com/showpost.php?p=1787829&postcount=590
That was beautiful, can't wait to see/hear what the devs do with all this audio hardware.
So, to conclude, most of the power on the audio block isn’t available to developers, but reserved for Kinect processing, but that doesn’t include SHAPE. Shape is totally there for developers.
Bkilian made the post you linked to on the 24th of September, but just a day before, he was more specific and pointed out that as far as he was aware, devs don't have access to the 4 DSP cores, but he stated that they have full access to the fixed function hardware. SHAPE is all fixed function hardware. Six blocks of fixed function hardware, to be exact.
Bkilian gives the reason why SHAPE was created.
http://beyond3d.com/showpost.php?p=1764893&postcount=223
Any numbers I could give you would be irrelevant. Games are not about FLOPS. SHAPE was created in order to give developers more flexibility in where they want to spend development time. Do you a) spend a bunch of time converting your audio engine over to SHAPE, and then save time on optimizing your physics engine, or b) leave the audio on CPU and spend a bunch of time optimizing the hell out of your physics engine? In the end, you will optimize and fiddle with the engines exactly the amount required to get the effect the designers want.
I like to use the example of the HD DVD player on the 360, since I have personal experience with it. When we started, we were using 100% of the 360, and getting 1 fps on a 24fps H.264 stream. When we finished, we were using 100% of the 360, and getting 24fps on a 24fps H.264 stream. Could we have optimized even more and gotten 30 or 60fps on that stream? Yes. Did we? No. It performed to spec, so we stopped right there.
In theory, the audio hardware in the One can produce results that could not be replicated by the entire 360 CPU. But what about good enough? If you don't use a polyphase SRC, and just go for the standard linear SRC? Now your audio is not quite as good (but you'd be hard pressed to tell most of the time) and you've reduced your CPU requirements a ton.
SHAPE was a free addition to the Xbox One audio block that Microsoft was lucky enough to get from the chip designers that did the 4 DSP cores designed to handle the Kinect processing.
http://beyond3d.com/showpost.php?p=1765802&postcount=277
Not bad. The Audio processor was originally devised to be able to offload Kinect Audio processing, and the chip designers came to the audio team and said "We have a bunch of extra transistors we can throw in for free, what would you like them to do?" or something close to that. The SHAPE block was the result of that conversation.
It's unmistakable at this point that devs have full access to SHAPE. It's the rest of the audio block, where the most powerful DSP cores reside, that they don't have full access to
