i heard that some guys are trying to make a program which is processed by a nvidia card
they said that the best nvidia card has the power of a 10ghz intel processor
is there a way to make a vsti plugin which uses a nvidia gpu for rendering audio data ?
or better a vsti wrapper which uses the gpu ?
vsti´s for grafic cards ?
-
- KVRist
- Topic Starter
- 100 posts since 18 Mar, 2002 from Austria
-
- Pick Me Pick me!
- 9696 posts since 12 Mar, 2002 from a state of confusion
see this is what im talking about!! 
out of the blue, we had this discussion last week
the thread was titled "Use the GPU" im sure if you searched for it you'd find all the info you desire on this subject
out of the blue, we had this discussion last week
the thread was titled "Use the GPU" im sure if you searched for it you'd find all the info you desire on this subject
-
- KVRAF
- 1907 posts since 29 Oct, 2003
that would be nicer. even more nicer if it would be featurepacked like energyXTor better a vsti wrapper which uses the gpu ?
hmm - what about fireGL's ?
-
- KVRAF
- 6478 posts since 16 Dec, 2002
sometime last spring I bumped into one of the early spec sheets of BrookGL, which is what this whole 'C on CG processors' is all about. I (half jokingly) talked to some people about starting a more audio oriented research about this here at my university.
Of course the brook people at stanford are way of head of the game so it would be useless at this point. Let them get the intial speedbumps out of the way. And they will as they have proper funding as well.
A couple of interesting point came up though.
CG cards of today are not *yet* very suitable for audio algoritm processing. There's not enough branching capabilities (ie. shaders are not programmable enough). Audio code is also kind of hard, at least very timeconsuming, to parallelize to the extent required by Brook.
Next generation graphics cards, coming out this spring, are going to be much better suited for audio needs.
By the way, BrookGL will work on all main platforms and API's. Yes, that means Radeons, Geforce FX's, and OpenGL/DirectX alike. And as Brook is just an extension to C++ code, it'll be generally quite painless for developers to adapt. Think of it kind of like SSE extensions... Not like propietary UAD/TC card stuff. Although there will definetely be similar latency issues, at least with the current architecture (PCI[express]/AGP).
So yes, audio plugins on graphics cards is definetely the future. Expect the first attempts in late 2004.
The benefits of this approach come, of course, not just for audio. We are definetely living interesting times indeed.
As bus architectures speed up, I'm sure all this will in time generate a market for more general DSP accelerators.
Of course the brook people at stanford are way of head of the game so it would be useless at this point. Let them get the intial speedbumps out of the way. And they will as they have proper funding as well.
A couple of interesting point came up though.
CG cards of today are not *yet* very suitable for audio algoritm processing. There's not enough branching capabilities (ie. shaders are not programmable enough). Audio code is also kind of hard, at least very timeconsuming, to parallelize to the extent required by Brook.
Next generation graphics cards, coming out this spring, are going to be much better suited for audio needs.
By the way, BrookGL will work on all main platforms and API's. Yes, that means Radeons, Geforce FX's, and OpenGL/DirectX alike. And as Brook is just an extension to C++ code, it'll be generally quite painless for developers to adapt. Think of it kind of like SSE extensions... Not like propietary UAD/TC card stuff. Although there will definetely be similar latency issues, at least with the current architecture (PCI[express]/AGP).
So yes, audio plugins on graphics cards is definetely the future. Expect the first attempts in late 2004.
The benefits of this approach come, of course, not just for audio. We are definetely living interesting times indeed.
As bus architectures speed up, I'm sure all this will in time generate a market for more general DSP accelerators.
