how long until...

[NeoGen]

How long do you think it will take until someone starts looking into soundcards for DC'ing as well? :)

The Sound Blaster X-Fi audio processor packs in over 51 million transistors that churn out over 10,000 MIPS of power!

http://www.soundblaster.com/products/x-fi/
http://www.soundblaster.com/products/x-fi/technology/xfiaudio/

[AMDave]

Probably not long looking at that number.
It depends on whether or not the SPU manufacturers will allow it by releasing their chip level command sets, i guess? Someone would have to build a compiler to enable it to do your bidding.

A bit of a scout around found this
http://www.oreilly.de/catalog/multilinux/excerpt/ch14-01.htm
if you want to start playing with it.
You can write and read to the devices, but it is not clear to me whether you can drive the SPU to perform any operations other than what it was designed for based on the binaries already available.

I wonder if the sequencer could be modified to perform bit transformations?

hmmm.
I wasn't able to find anything to suggest that this has been tried yet.
Anyone else know of any text on this topic?

[AMDave]

This online article which leads up to the intorduction of a branded product is quite interesting.
http://edn.firstlightera.com/EN/Microsites/1/Tensilica/Low-PowerLow-OverheadHigh-FidelityDigitalSoundforSOCsTensilicasHiFi2AudioEng ine.htm

Evidently sound DSPs (Digital Signal Processor) became so specialised because CPUs couldn't process the codecs fast enough. However, this has isolated them as a processor that generally hangs around as a limp appendage waiting for something to do. As a result, naturally enough, there are moves afoot to move sound DSPs back onto enhanced RISC platforms, which I expect would ultimately enable the sound DSP to be used for other purposes as well as processing sound. Makes perfect sense really.

On further reading of the product paper
http://edn.firstlightera.com/ERA/ResourceGallery/ResourceHandler.ashx?b034cb3e-b8b7-4ae0-9176-c2e80f489a4d;HiFi2%20White%20Paper%20v2.3.doc
it seems that effectively what this group have done is build sound extensions back into the RISC-CPU .... which means that basically because the sound DSP IS a CPU, there is no reason why it can't also run the OS, which begs the next step ... eliminating sound cards altogether ! ! !

Do you think AMD could build this into the chipset so it could add the power of a RISC-sound DSP onto the Hypertransport channel?
(sounds a bit like the Power-PC though, huh?)
Hey! There would be a whole new market out there for "cooling" equipment for your sound DSP! hehe

Actually I think these people are onto something.
The sound card has for a long time stared back at me as an over-Marketing success:
"Like, Dude! we totally sold you something you hardly ever need to use. Ha Ha Ha!"

Combining the DSP function back into a RISC-CPU (or even a CISC-CPU) that could double as a co-processor would make me go out and buy it because I'd actually be adding something to my system that would improve it's performance rather than just help it go "BEEP!" better.