However, it was shown on a QuadroFX board of course, and it is anybody's guess whether Adobe and/or nVidia will allow non-QuadroFX boards to take advantage of this acceleration. There is some indication that CS5 will bring heavy changes in this area, with a lot more GPU-accelerated effects in the timeline (basically just about every significant effect), at least that is what they were showing at IBC, on a nVidia platform. Even nVidias solution is only a encoder (on the Quadro-line) plugin which does nothing ATM to offload the RT editing in the timeline, and even what it does accelerate (basically H.264 encoding) is rather limited. If you are editing HD H.264 material, then it is highly likely that the CPU is the problem, and for now you only have something like the Divide Frame GPU decoder that could help with offloading the decoding on a GPU. However, whether in your specific case there will be a noticeable improvement, depends on if the GPU-accelerated functions are really the bottleneck for your projects. For now, I believe, everything is internally implemented via DirectX, no CUDA or OpenGL (but OpenGL is used for AfterEffects, so if you also use that.). Generally speaking, you will see an increase in interactivity with a better card - yours is I believe completely unsupported, and in any case you need to have a 512 MB card for best performance. There are also third-party plugins, like Magic Bullet Looks, which are pretty heavily GPU accelerated. There is a certain amount of GPU-acceleration going on in PPro CS4, and not just the few GPU effects which are directly computed on the GPU, but also in terms of (re)scaling the preview buffers, and even sources.
0 kommentar(er)
