
Now that we indeed the subject already
in conjunction with Final Cut Pro had touched on X, we want to push at this point to some interesting benchmarks. Thus studio1productions different, sometimes only unofficially supported Nvidia graphics cards with the Mercury Engine by Adobe Premiere Pro tried (see item 9 in the linked document). With the amazing result that almost no speed difference was detected between maps. Whether Celt 96 cores in a favorable GT240 or 480 cores in an expensive GTX 570, the difference in rendering speed enabled GPU is not fast. This is also consistent with our experience, and supports the thesis that most of the time consumed in getting the image data from main memory into the GPU. And here are the cards almost indistinguishable. The current bottleneck seems thus not the GPU power but the data transfer to the graphics card to be. We want to systematically investigate the issue once, but still work out the way we do the best ...