CES 2014: AMD Kaveri Tidbits

"It's the most wonderful time of the year..." OK, aside from the Polar Vortex as of late, for all us tech junkies it's a special time of year because CES is upon us! This year AMD has unveiled their new chip, Kaveri, and though there was not a lot of new info (no benchmarks publicly available yet), there are a few particularly interesting details.

A while back AMD joined The Document Foundation, offering to help refactor the code for LibreOffice Calc. Essentially what the folks at LibreOffice have accomplished since then is to leverage the hardware found in AMD systems (in particular, the GPU, or in the case of Kaveri, the APU) to do some very heavy lifting. Exactly how much the new hardware and code improved performance was revealed at CES:

"Now that Kaveri offers a proper HSA stack, and can call upon 12 compute cores to do work, applications that are designed (or have code paths) to take advantage of this should emerge.  One such example that AMD are willing to share today is stock calculation using LibreOffice's Calc application – calculating the BETA (return) of 21 fake stocks and plotting 100 points on a graph of each stock.  With HSA acceleration on, the system performed the task in 0.12 seconds, compared to 0.99 seconds when turned off."

Source: AMD’s Kaveri: Pre-Launch Information, by Ian Cutress (anandtech.com). 

(slide source)

Looks like progress to me.

The other interesting thing of note is how AMD is shipping Kaveri, (and certain discrete GPU offerings), with TrueAudio. True Audio? Yeah, that's what I thought too. We know it is being shipped with chips in the Sony PS4. Ryan Smith did a nice article on TrueAudio over at Anandtech and will give you a good idea of what it's all about. But what caught my attention was a little mention that the folks at Nuance have begun working with AMD TrueAudio for their voice recognition software.

"Software support is improving with applications from companies like Adobe, The Document Foundation, and Nuance.  These cover HSA applications and in Nuance’s case, using the TrueAudio portion to clean up and accelerate voice recognition."

Source: AMD CES 2014 Presentation: Kaveri Goes Official, by Josh Walrath (PC Perspective)

In a nutshell, it looks like TrueAudio will enable audio processing by the GPU (or the GPU part of the APU)¹ which will allow voice recognition software to work well, despite being deployed in noisy places like an office:

"Other demonstrations included HSA/TrueAudio being used by Nuance to clean up audio streams to the point where they could do speech/voice recognition even in noisy environments."

Source: AMD Press Conference Highlights, by Jarred Walton (anandtech.com)

Initially, I had assumed that "speech/voice recognition," referred to Nuance's flagship product, Dragon Naturally Speaking. But there is more to this story. According to Clarice Simmons, AMD has been working with Nuance to develop voice recognition software (based on Nuance's Voice Biometrics) that would go together with AMD's Face Login software. She hints that TrueAudio might be employed by Nuance to clean up the audio input with regards to voice recognition:

"For example, AMD TrueAudio technology  (the sophisticated ACP or Audio Co-Processor) built into AMD’s fourth generation APU “Kaveri” could be used to clean up the voice input for even greater discretion and accuracy in voice recognition."

Source: Could your face—and your voice—be the key to your fortune?, by Clarice Simmons, Senior Marketing Manager at AMD.

This would make sense, as the problem of dictating to Dragon in a noisy office would also be a problem when entering a voice recognition phrase to gain access to your computer.

In theory, the voice recognition would add a level of security to the process; face recognition alone could have some weakness:

"You can use AMD Face Login to log into Windows and even some of your favorite websites. This can be really convenient, but some people worry that someone using a high definition picture or even video can fake out the software. They needn’t worry, we are incorporating additional security to be on the safe side… Voice authentication."

Just be sure you don't store anything like those pesky Blackbriar files on your computer. =)

With regard to dictating in a noisy environment, I think it would be nice to see Nuance harness the AMD hardware technology for Dragon as well. As someone who has suffered from Tennis Elbow, I can attest to the worth of a program like Dragon Naturally Speaking. There was a time where I had to rely heavily on that program to get my desk work done, and it worked admirably well.

I have been told by a reliable source that the vast majority of improvements to Dragon in recent years have been connected to the hardware. In other words, the more hardware speed/processing power improves, the better Dragon works. That said, it appears that hardware advances with respect to raw processing power are slowing, so the gains from faster hardware will also slow. But an area where Dragon can definitely improve is its ability to get good audio input. If AMD TrueAudio can help to this end, it will be a welcome addition by all who rely on this software.

Notes:
^{____________________________________}

1. Granted, the difference between CPU & GPU is becoming harder to distinguish, as AMD brings the two units closer together.