Kommentare

FrigidNDEditing schrieb am 05.03.2009 um 05:59 Uhr
and it's basically just for transcoding - so I don't see where it will benefit us in the initial render either.

Dave
farss schrieb am 05.03.2009 um 06:37 Uhr
"I don't think Cuda is required for 3D acceleration. OpenGL has been around for a long time."

And there's the problem. What's being discussed here is not really "3D acceleration" at all. That seems to be where all the confusion originates from because what's being discussed is rendering bitmaps not rasterising vectors.

What CUDA offers is the ability to use the large number of pipeline processors to run calcs that can accelerate encoding tasks for example by assigning one pipeline per macroblock for H.264 encoding/decoding. That too is probably of no benefit for transforming bitmapped planar surfaces in 3D,

However I have seen CUDA used to accelerate some bitmap operations such as zooming into an enourmous image in Photoshop so maybe, just maybe, it could be used to some advantage in the kind of image translation being discussed here.

Bob.

rmack350 schrieb am 05.03.2009 um 14:54 Uhr
True. I see what you're saying. The picture I had in mind was mapping video onto a vector shape and then moving/flipping it around. You're right that bitmap rendering still has to be done and CUDA ought to be able to help with that.

When we were first getting demos of Axio cards on PPro CS2 they were showing us this sort of thing - how you could take a pip and roll and flip it all over the place. Seemed like you could at least get a good idea of the motion and then render it afterwards, if I remember correctly. The thing about these setups was that an Axio rig also required a beefy video card to do this sort of processing, and at that time this certainly wasn't CUDA or Stream.

In any case, SCS has never given much of a hint that they'd support any sort of hardware acceleration. The user understanding is that Vegas is hardware neutral. If that were to hold true then I think you'd see Vegas start to use hardware acceleration a few years (decades?) after it became a standard part of the Windows OS.

CUDA get's things a bit closer because it's a system to do fairly generalized processing. We'd all love to see third party plugins and codecs using it but I kind of doubt that SCS would tie themselves exclusively to NVIDIA.

Rob Mack
John_Cline schrieb am 05.03.2009 um 23:18 Uhr
I believe that I read somewhere that ATI and nVidia are getting together to come up with a standard for GPU acceleration programming.
Himanshu schrieb am 06.03.2009 um 01:49 Uhr
Not sure if you were referring to OpenCL which I mentioned in this thread? The group of vendors proposing/supporting that include AMD, NVIDIA and Apple among many other noteable companies.