Xbox One (Durango) Technical hardware investigation

[mrcorbo]

And your receiver is just going to throw a 24Khz low pass filter on the audio and output it that way, since allowing the ultrasonic frequencies through will a) cause your speakers to alias and reduce audio quality, and b) if they can be reproduced, damage your hearing.

After reading the article mczak posted and seeing it reinforced here, apparently it CAN hurt. I wasn't before aware that trying to reproduce the ultrasonic frequencies actually could affect the reproduction of the frequencies you could hear. That certainly changes things. If the best you can hope for is that it has no effect than it really is pointless.

Yeah, but 96 kHz is bigger ...

Being overly-reductive and dismissive doesn't tend to promote good discussion. There's already plenty of posters having a negative effect on the quality of the discourse here lately. You don't need to be one of them.

My earlier position was based on my general preference for reproducing source material with as little modification as possible. Not some ignorant belief that bigger is better. Now that I know that in this case that can actually be harmful, I'm quite willing to admit I was wrong and happy to have learned something I didn't know before. That last being one of the main reasons I post here.

 

[bkilian]

After reading the article mczak posted and seeing it reinforced here, apparently it CAN hurt. I wasn't before aware that trying to reproduce the ultrasonic frequencies actually could affect the reproduction of the frequencies you could hear. That certainly changes things. If the best you can hope for is that it has no effect than it really is pointless.

Yes, unfortunately, just because you can't hear a sound doesn't mean it can't hurt you. My aforementioned ultrasonic array outputs a 120dB (at 1ft) carrier wave, completely silently. You can actually blow your ears out without ever having realized there was audio there.

High definition audio is a very tricky subject, simply because it is so subjective, and some people insist they can hear the difference between 48Khz and 96Khz. Of course, like Uri Geller, their miraculous powers disappear completely under ABX testing, but then they just insist that their equipment is better than the testing equipment, or they were stressed, or the environment was too noisy, or any excuse to not have to admit that they paid a ton of money for snake oil.

Here's the cool part. _they're right_. They _can_ hear a difference (if they know the difference is there). Tests using non-ABX, with brain scanners, show that their brain reacts differently to (what they consider) the better audio, in effect, they enjoy that audio more because they "know" it's better.

 

[Scott_Arm]

Being overly-reductive and dismissive doesn't tend to promote good discussion. There's already plenty of posters having a negative effect on the quality of the discourse here lately. You don't need to be one of them.

My earlier position was based on my general preference for reproducing source material with as little modification as possible. Not some ignorant belief that bigger is better. Now that I know that in this case that can actually be harmful, I'm quite willing to admit I was wrong and happy to have learned something I didn't know before. That last being one of the main reasons I post here.

Fair point. The comment was not meant to be directed at you. I wasn't thinking about anyone in the thread. I was just joking with bkillian, but I can see how it could be taken the wrong way. My apologies.

 

[(((interference)))]

The SHAPE audio block and processing cores are really quite powerful (it's probably the closest thing in the actual hardware to the silly 'special sauce' claims).

Theoretically, they can free the main CPU from processing audio entirely - which is quite a benefit given games can use as many as 3 of the 6 CPU threads on 360 for audio.

Nevermind it also handling Kinect MEC and voice recognition which previously consumed CPU time from the game as well (as bkilian said above).

The SHAPE audio block alone is more than 100GF (though it can't really be counted that way since it's not general purpose FLOPS). The alpha kits had an extra 8 CPU cores to emulate the sound block but even that wasn't enough processing power so they never implemented the emulation
That's also where some early rumours of Durango having a 16 core CPU came from.

I don't think the PS4 has similarly capable fixed function audio hardware, so this helps compensate for the extra CPU core that's reserved on Durango.

 

[Scott_Arm]

Yes, unfortunately, just because you can't hear a sound doesn't mean it can't hurt you. My aforementioned ultrasonic array outputs a 120dB (at 1ft) carrier wave, completely silently. You can actually blow your ears out without ever having realized there was audio there.

High definition audio is a very tricky subject, simply because it is so subjective, and some people insist they can hear the difference between 48Khz and 96Khz. Of course, like Uri Geller, their miraculous powers disappear completely under ABX testing, but then they just insist that their equipment is better than the testing equipment, or they were stressed, or the environment was too noisy, or any excuse to not have to admit that they paid a ton of money for snake oil.

Here's the cool part. _they're right_. They _can_ hear a difference (if they know the difference is there). Tests using non-ABX, with brain scanners, show that their brain reacts differently to (what they consider) the better audio, in effect, they enjoy that audio more because they "know" it's better.

That's actually pretty fascinating. I've always wondered if a lot of the guys in the high-end audio stores knowing their selling some stuff that's a scam or if they actually believe that it's better.

Also, if I can ask, how significant is the processing required for the noise/echo cancellation on Kinect? I'd imagine it's not too much because they managed to fit voice control into Skyrim post-release as well as into Mass Effect 3.

 

[mrcorbo]

Fair point. The comment was not meant to be directed at you. I wasn't thinking about anyone in the thread. I was just joking with bkillian, but I can see how it could be taken the wrong way. My apologies.

Understood. Thanks for the clarification.

 

[expletive]

The SHAPE audio block and processing cores are really quite powerful (it's probably the closest thing in the actual hardware to the silly 'special sauce' claims).

Theoretically, they can free the main CPU from processing audio entirely - which is quite a benefit given games can use as many as 3 of the 6 CPU threads on 360 for audio.

How many jaguar cores is equivalent to 3-6 xenon threads?

 

[Nisaaru]

High definition audio is a very tricky subject, simply because it is so subjective, and some people insist they can hear the difference between 48Khz and 96Khz. Of course, like Uri Geller, their miraculous powers disappear completely under ABX testing, but then they just insist that their equipment is better than the testing equipment, or they were stressed, or the environment was too noisy, or any excuse to not have to admit that they paid a ton of money for snake oil.

Here's the cool part. _they're right_. They _can_ hear a difference (if they know the difference is there). Tests using non-ABX, with brain scanners, show that their brain reacts differently to (what they consider) the better audio, in effect, they enjoy that audio more because they "know" it's better.

I'm curious if you did test the effect of different digital filters too in your research? I can still remember that during mid to late 90s there was the "high end theory" that the effect of the steep digital filter necessary with 44.1khz was noticeable by the human ear. So some CD/DAC companies added different filter modes to their products.

 

[MrFloopy]

Just curious,

What would the relative die sizes be for the xenon as opposed to the sound IC described in the leak?

 

[bkilian]

That's actually pretty fascinating. I've always wondered if a lot of the guys in the high-end audio stores knowing their selling some stuff that's a scam or if they actually believe that it's better.

Also, if I can ask, how significant is the processing required for the noise/echo cancellation on Kinect? I'd imagine it's not too much because they managed to fit voice control into Skyrim post-release as well as into Mass Effect 3.

Depends on when you link the libs. We released multiple versions of the system, each version using less resources. The MEC algorithm uses about 1-2ms of CPU time on a single thread out of each 16ms frame. I'm not sure how heavyweight the voice reco is. Memory usage is on the order of a few megs.

 

[Cjail]

Then, why is it different?

There are two articles about Drunago's memory.

-The first one shows the Durango memory clients with the maximum theoretical bandwidth available in every path.

-The second one it's just an example of Dutago's memory system and it simply shows how Dutago would work under a "hypothetical" load at certain conditions.

 

[Gubbi]

I'm curious if you did test the effect of different digital filters too in your research? I can still remember that during mid to late 90s there was the "high end theory" that the effect of the steep digital filter necessary with 44.1khz was noticeable by the human ear. So some CD/DAC companies added different filter modes to their products.

When sampling an analog signal at sampling frequency f, the spectrum is folded around f/2. This means you need to remove every frequency above f/2 before sampling (otherwise a 44KHz frequency also appears as a 100Hz frequency).

If you're sampling @ 44.1KHz, you need a steep low pass filter, like a high order Butterworth or Chebyshev filter to cut frequencies above f/2 (or a little lower, 20KHz). High order analog filters has significant phase shifts towards the cut off frequency, this might be audible.

In the digital domain you can build filters with low (or no) phase distortion, but at higher computarional cost and with higher delay. Delay has zero significance in a recording situation, so by using a much higher sampling frequency, say 96KHz, one can push the cut off frequency of the analog filter well into the ultrasonic part of the spectrum with zero phase distortion below 20KHz. Then when the source is sampled apply a zero phase distortion low pass filter with a cut off frequency of 20KHz to it and then downsample it. You now have a 44.1KHz digital recording without any phase distortion.

In playback situations, you have a low order reconstruction filter, because you can't hear ultrasonic content anyway, and unless you're bkilian your speakers have very low sensitivity to high frequency contents anyway. The only other place filtering is used to any extent is in your equalizer and your room correction DSP.

Cheers

 

[DieH@rd]

Is DDR3 memory supporting clamshell mode? Quick Google search is not giving me anything.

 

[itsmydamnation]

How many jaguar cores is equivalent to 3-6 xenon threads?

depends by what measure, anywhere from 1 to 3 would be a pretty good ball park(assuming something like 2ghz clock).

 

[Xenio]

The SHAPE audio block and processing cores are really quite powerful (it's probably the closest thing in the actual hardware to the silly 'special sauce' claims).
.

It's funny that as we're reaching the reveal day, Xbox Infinity seems to be a dual SOC machine (xbox360 BC soc) and special sauce (SHAPE audio block), if the final frequency of GPU/CPU will be different from 1.8/0.8 GHz then all the three things I though months ago was all right

 

[Love_In_Rio]

The SHAPE audio block and processing cores are really quite powerful (it's probably the closest thing in the actual hardware to the silly 'special sauce' claims).

Theoretically, they can free the main CPU from processing audio entirely - which is quite a benefit given games can use as many as 3 of the 6 CPU threads on 360 for audio.

Nevermind it also handling Kinect MEC and voice recognition which previously consumed CPU time from the game as well (as bkilian said above).

The SHAPE audio block alone is more than 100GF (though it can't really be counted that way since it's not general purpose FLOPS). The alpha kits had an extra 8 CPU cores to emulate the sound block but even that wasn't enough processing power so they never implemented the emulation
That's also where some early rumours of Durango having a 16 core CPU came from.

I don't think the PS4 has similarly capable fixed function audio hardware, so this helps compensate for the extra CPU core that's reserved on Durango.

i supposse the competition will have a chip capable of processing 5.1 multichannel sound. With that is enough.

 

[Xenio]

i supposse the competition will have a chip capable of processing 5.1 multichannel sound. With that is enough.

what do you mean with "capable of processing 5.1 multichannel sound" ?
even a cheap audio chip inside a dvd player can process and output 5.1 but it don't fit well in a next gen console
how many voices, with what effect and with what quality?
and I suppose that comparison with competition are not allowed here

 

[MrFox]

These are amazing specs for a sound chip, it's in the "what-more-could-you-possibly-need" category :smile:

The SHAPE audio block alone is more than 100GF (though it can't really be counted that way since it's not general purpose FLOPS). The alpha kits had an extra 8 CPU cores to emulate the sound block but even that wasn't enough processing power so they never implemented the emulation
That's also where some early rumours of Durango having a 16 core CPU came from.

I'm surprised they would try the emulation with a CPU? I would have thought this kind of code to be perfect for a GPU. Any idea what would make the CPU a better match?

 

[liolio]

With such an audio chip I would like to be able to tun Music/Guitar software on the system (AMpfarm and the like)
Could be interesting if those apps are released as metro apps and can run on the system.
Durango seems like a pretty well rounded system to me.

 

[Love_In_Rio]

what do you mean with "capable of processing 5.1 multichannel sound" ?
even a cheap audio chip inside a dvd player can process and output 5.1 but it don't fit well in a next gen console
how many voices, with what effect and with what quality?
and I suppose that comparison with competition are not allowed here

Well, Xbox 360 was capable of real time 5.1 sound processing with as much as 3 hardware threads. So, a chip capable of processing that would be enough for most people.