Intel Broadwell gets integrated DSP

[Fjodor2001]

http://vr-zone.com/articles/meet-intels-hardware-level-siri-killer/60323.html

Intel’s updated Smart Sound Technology DSP found in 2014′s Broadwell promises to outsmart Siri in every way possible. The catch? It’s already partially here in some Bay Trail configurations.
[...]
Siri’s (and other voice recognition solution’s) fatal flaw is that in order to be work at least semi-accurately the user needs to be speak to it a direct voice: louder than average; focused and not natural.

​

Adding to this lack of natural feeling is that voice recognition solutions today are “push-to-talk”. They have to be triggered. On the iPhone this means hitting the Home button to get Siri to ask “what can I help you with?” or hitting the microphone button in Android to get it to patiently listen for your request.

Intel is planning on changing that with its refreshed Smart Sound Technology (SST) DSP that’s currently in some Bay Trail configurations, and will be wholly integrated with 2014’s Broadwell. It promises to offer a more natural voice recognition system than current ecosystem offerings, without being as taxing on the CPU — as Siri and Google’s voice actions are — as it will bypass it entirely.

 

[liahos1]

http://vr-zone.com/articles/meet-intels-hardware-level-siri-killer/60323.html

this is really interesting if it works. I was never a huge fan of siri on my 4s and i dont use anything like siri on my htc phone now. If i can get an htpc to listen to what i have to say and it halfway does what i ask it to do ("load crysis 4" ) its a great feature.

 

[Exophase]

Holy crap does Intel know how to play up marketing hype. Everybody in the SoC world has a DSP already. It was inevitable that Broadwell would get one too, since Intel has gradually been moving their desktop/laptop chips to full SoC integration. But all the crap about beating Siri and being able to talk to it with "Hello computer" and whatever is pure OS-level or higher software, not hardware magic or DSP firmware, and nothing that competitors can't do. I knew that before even looking at their little diagram which clearly reaffirms it, but that just goes to show that the article is pretty misleading in trying to spin this as a hardware advantage.

I haven't used Siri, but if it's really supposed to be superior to Android's voice recognition then it must be pretty good because I've gotten consistently good voice recognition on my Nexus 4. I think Intel is just talking trash here.

 

[PPB]

Holy crap does Intel know how to play up marketing hype. Everybody in the SoC world has a DSP already. It was inevitable that Broadwell would get one too, since Intel has gradually been moving their desktop/laptop chips to full SoC integration. But all the crap about beating Siri and being able to talk to it with "Hello computer" and whatever is pure OS-level or higher software, not hardware magic or DSP firmware, and nothing that competitors can't do. I knew that before even looking at their little diagram which clearly reaffirms it, but that just goes to show that the article is pretty misleading in trying to spin this as a hardware advantage.

I haven't used Siri, but if it's really supposed to be superior to Android's voice recognition then it must be pretty good because I've gotten consistently good voice recognition on my Nexus 4. I think Intel is just talking trash here.

TrueAudio seems to be capable of doing the same thing for the next gen consoles. Certanly, its an added value, but it never ceases to amaze me how polar reactions can get from the same feature, just on depending who is the company behind it.

 

[Revolution 11]

A DSP in a SoC is quite different from DSPs in GPUs. The former is common, maybe even required for certain smartphone functions, the latter is unheard of until TrueAudio.

Putting better than mainstream audio and video capabilities together into a single package interests me. Who wants both a sound card and GPU? I hope Nvidia gets their act together and also adopts some form of audio processing.

 

[VirtualLarry]

I hope Nvidia gets their act together and also adopts some form of audio processing.

Interesting note: The first NVidia graphics chip, the NV-1, also had audio capabilities as well. (As well as Sega Saturn joystick ports)

 

[BallaTheFeared]

Awesome, my next i5-6670k will have an iGPU and a iAudio!

With 20% more IPC, lulz!

Should be able to boost my Mantle increased performance with Virtu by another 10-20%, and enhance my AMD true audio performance with 4D sound!


Seriously, so many gimmicks, just give me performance

 

[Techhog]

Awesome, my next i5-6670k will have an iGPU and a iAudio!

With 20% more IPC, lulz!

And you'll need custom watercooling to get to 4GHz.

 

[Revolution 11]

Long before my interest in PCs started but they need to go back to that idea. Since there is practically no competition or innovation going on in standalone audio cards.

 

[Shivansps]

Holy crap does Intel know how to play up marketing hype. Everybody in the SoC world has a DSP already. It was inevitable that Broadwell would get one too, since Intel has gradually been moving their desktop/laptop chips to full SoC integration. But all the crap about beating Siri and being able to talk to it with "Hello computer" and whatever is pure OS-level or higher software, not hardware magic or DSP firmware, and nothing that competitors can't do. I knew that before even looking at their little diagram which clearly reaffirms it, but that just goes to show that the article is pretty misleading in trying to spin this as a hardware advantage.

I haven't used Siri, but if it's really supposed to be superior to Android's voice recognition then it must be pretty good because I've gotten consistently good voice recognition on my Nexus 4. I think Intel is just talking trash here.

The thing is, no much people cares about voice recognition, i whould not call if hype because is was not marketed as TrueAudio was, that is hype at a AMD common level.

I just dont care about DSP, and i bet a lot of people is with me on that one, im agasint any "brand only" thing, we dont need that. So this Intel announcement could be usefull on mobile, but i just hope they keep it away from desktop.

 

[Techhog]

Holy crap does Intel know how to play up marketing hype. Everybody in the SoC world has a DSP already. It was inevitable that Broadwell would get one too, since Intel has gradually been moving their desktop/laptop chips to full SoC integration. But all the crap about beating Siri and being able to talk to it with "Hello computer" and whatever is pure OS-level or higher software, not hardware magic or DSP firmware, and nothing that competitors can't do. I knew that before even looking at their little diagram which clearly reaffirms it, but that just goes to show that the article is pretty misleading in trying to spin this as a hardware advantage.

I haven't used Siri, but if it's really supposed to be superior to Android's voice recognition then it must be pretty good because I've gotten consistently good voice recognition on my Nexus 4. I think Intel is just talking trash here.

Google Now has superior voice recognition compared to Siri. Anyone who says otherwise is either Apple or a brainwashed Apple fanatic.

 

[lothar]

How is this different from Motorola's "always listening" device like the Moto X?
Motorola's voice recognition is not “push-to-talk”. And it doesn't have to be triggered either.

 

[WhoBeDaPlaya]

 

[Sable]

LOL. Exactly what I was thinking.

 

[SammichPG]

Holy crap does Intel know how to play up marketing hype. Everybody in the SoC world has a DSP already. It was inevitable that Broadwell would get one too, since Intel has gradually been moving their desktop/laptop chips to full SoC integration. But all the crap about beating Siri and being able to talk to it with "Hello computer" and whatever is pure OS-level or higher software, not hardware magic or DSP firmware, and nothing that competitors can't do. I knew that before even looking at their little diagram which clearly reaffirms it, but that just goes to show that the article is pretty misleading in trying to spin this as a hardware advantage.

I haven't used Siri, but if it's really supposed to be superior to Android's voice recognition then it must be pretty good because I've gotten consistently good voice recognition on my Nexus 4. I think Intel is just talking trash here.

The sad truth is that voice recognition is crap outside of english, you usually have to talk like a robot and half of the time the thing fails because of accented letters (èàùò).
I feel completely dumb speaking in public to my phone.

Text to speech is funny as well, android gps nav can't spell correctly so many street names and I bet apple maps is just as bad.

 

[Nothingness]

It's hilarious to see Intel "inventing" what SoC have been doing for 15 years Or perhaps a better word is ridiculous rather than hilarious?

 

[liahos1]

It's hilarious to see Intel "inventing" what SoC have been doing for 15 years Or perhaps a better word is ridiculous rather than hilarious?

i thought these always on dsp's were relatively new for the consumer market. e.g. the motox and droids seem to have them in those "8 core" configurations. thats pretty good marketing too dont you think

 

[sm625]

Is intel doing this for consumers, or for the NSA?

 

[Revolution 11]

The NSA is one of their consumers. So both....

 

[mrmt]

But all the crap about beating Siri and being able to talk to it with "Hello computer" and whatever is pure OS-level or higher software, not hardware magic or DSP firmware, and nothing that competitors can't do.

So everyone has already a DSP that bypasses the CPU entirely?

 

[Exophase]

So everyone has already a DSP that bypasses the CPU entirely?

This doesn't even make sense. Bypasses the CPU entirely.. while doing what? If you think that this "Genie" stuff is running entirely on the DSP then you failed hard at reading the block diagram.

The DSP is there to do what DSPs do - signal processing. Meaning, in this case, sampling, filtering, decimation, maybe some packetizing stuff. And yes, the same thing everyone else is already doing.

 

[TechFan1]

"This doesn't even make sense. Bypasses the CPU entirely.. while doing what?" It listens for voice wakeup command, while cpu is in connected standby. Not the only soc with this feature, but it may quickly become a necessary feature for any chipmaker wanting to sell mobile chips.

 

[krumme]

Intel needs absolutely excellent dsp in their cpu moving forward. Look at eg. Qualcomm s800 and eg its 4k video recording (i assume its handled by the dsp). Its good Intel started. Now they need to get it 100% right with 14nm Atom the be serious in the phone market.

I think its fine we get it in other products its hardly many mm2 or what?

 

[Exophase]

"This doesn't even make sense. Bypasses the CPU entirely.. while doing what?" It listens for voice wakeup command, while cpu is in connected standby. Not the only soc with this feature, but it may quickly become a necessary feature for any chipmaker wanting to sell mobile chips.

I don't know what makes you think this "Hello Computer" processing code is running on the DSP. Assuming that the DSP is really even making a decision to wake up the CPU at all it could be on a much simpler noise profile than any kind of language recognition.

You guys are really missing the bullet that talks about using the DSP for real-time mixing, decode, and post-processing. Using DSPs for high level language processing is a bad idea.