Free gpu verb

[the19thbear]

this is a really cool thing! .. if only i had the brains/education, i would for sure team up with someone and explore this idea further!!

[astroman]

yes, convolution is a very obvious application for the high-speed multi-channeled Ram of a graph card.
Yet it doesn't apply on all and everything...
... but it did apply to cr*ck WPA keys by brute force recently
so people start catching up with the somewhat obscure programming interface

cheers, Tom

[pollux]

I've never yet owned nor wanted a Nvidia based gfx card. Algorithms like those should be useable on any common and sufficiently powerful gpu available before solutions like that become an option to even think about. Until that time comes and stuff starts working nicely and generally compatible, I will continue using my preferred hardware 'verbs. Also I believe, true audio focused development will always be one or several steps ahead of the software designers when it comes to useability and professionality. And there are those who will always seek their own sound and use their own flavor of gear, not paying much attention to what is hot and new (like me, more or less), or the vintage lovers who won't so much as twitch a brow at some analogue hiss or narrow stereo imaging.
The way it is now, I don't really see it being more than a tryout of new options, with existing, but easily overestimated chances of turning into something common, let alone a standard.

In fact it's not the algo that introduces the "compatibility issue", but rather that NVidia are the only ones out there who published an API to code against the GPUs (CUDA).
The same algo can be applied to any GPU (well, there aren't that much out there either ), provided that you have the interfacing API.

[JoeKa]

Yes pollux, I know that it's not a matter of the algo itself, but the code it is formulated in. I just simplified the phrasing there, as not everyone has an idea of programming terminology.
However, ATI/AMD has recently introduced STREAM, which is their approach to an api similar to cuda. And in fact, it's even possible to feed a gpu non hardware implemented codes using the DX instruction set. An example here would be the fft3dgpu video denoising filter for avisynth, which I have been using before and with great success. I don't think spacial-temporal or wavelet based video denoising is a native function of any gpu, or is it?

[astroman]

this looks like a useful cross-ref for GPU based programming projects
http://www.gpgpu.org/cgi-bin/blosxom.cgi/index.html

cheers, Tom

[dawman]

Just demo'd some Acusticaudio IR's...........yeah well.
No better than a VST like GigaPulse. Maybe they'll make some higher quality early reflections, as these giant halls and sewers are....yawn, yawn,.......more of the same.
Why do people think an IR has to be of a Cathedral or Arena and other gigantic places which you will never hear a synth.
Quality reverb should hardly be noticed, that's when you know something good is being used.

Please help me get my Bricasti Model 7
http://cgi.ebay.com/LEXICON-PCM91-DIGIT ... 240%3A1318

I have talkled my singer into finally breaking into her vault and kicking in a 1000 bucks on one w/ me. This is great news, as it is mine, she owes me bigtime, and I will provide her with a good DSP verb from the XITE-1 ( in a month or so ) and tell her how great she sounds, she'll never know the difference.. , while I apply the Model 7 to my new material.............ankyuvarymush.

[astroman]

regarding convolution: just installed the demo of Peavey Revalve
some parts of the devices are (apparently) physical modelling, some are convolution based
the modelling part seems pretty good so far, better than Guitar Rig, but the convolution stuff s*cks mucho.
Their spring reverb (IR based) revealed a fundamental fault: total lack of dynamic
in other words one would need at least a dedicated response for each significant level, while the virtual spring has to get along with a static sample.
If Irc handling of 'dynamic' is one of the core components of the Acoustic Audio method, but I'm not really sure.
The static way is just rubbish - probably the reason why it's often perceived as lifeless and artificial.

cheers, Tom

[Zer]

http://monalisa-au.org/

[Zer]

http://www-sop.inria.fr/reves/projects/GPUAudio/

But wasn´t there a project called bionicfx or sO? But that`s been canclled as far as I can see.

But loads of interesting sites if you are searching for gpu audio effects in google.

[valis]

As you guys have shown, Nvidia wasn't the first by any means to come up with the idea nor the first to bring something to market (UAD for example). In fact Nvidia has been touting number-cruching with their gpu's for a long time actually. Seems like it wasn't until the g80 GPU that they had a decent codebase out usable for real development purposes (CUDA), that was roughly 2 years ago (g80=8800GTX, 8800GTS 320MB/240MB). Now of course AMDATI is part of the "OpenCL" consortium, while Intel is working on their massively parallelized Larrabee which they aren't even really positioning as a gpu as much as competition for NVidia's "supercomputer on a card" high end CUDA platforms.

The guy doing BionicFx probably moved onto other things while waiting on for the market to get in gear...