
already almost irrelevant haunted today one of the coolest announcements in recent years by the Blender world: For the Blender compositing system there is a first OpenCL implementation. It already circulating www.youtube.com/watch?v=sVkDx_4GP5M&feature=player_embedded #! (A demo-video) in the network, showing a drastic speed increase after the program a test case even on a Core i7-920 and then to a sub-class graphics card (GT220) renders: example, the CPU needed for example for a clean Bokeh Blur
204 seconds, which for 70 € graphic card in 22 seconds now creates. Jeroen Bakker, one of the heads of the project calls seem indifferent to the problems still meet the current code uses compositor very much a shared memory model. However, this is the worst case for GPU applications that can work very quickly only if each thread can access its own local memory on. OpenCL this purpose, 25 percent of the graphics card memory to use it. With a 1GB card that is only 256 MB of RAM is used for the compositor, which quickly can become scarce at HD resolutions. But all this will get the project under control, the OpenCL support in future versions of Blender direct involvement. This is due among other things, already the use of multiple graphics cards on the master plan of the developer. This also means that the Compsiting system by the Nachstecken of graphics cards to accelerate can speed the. The first demo version runs only on 64 bit Linux and supports the following nodes Defocus, blur node (in bokeh mode), Color Balance , Lens Distortion, brightness, rotate and Tonemap. And while we have written this news, seems to have a programmer the corresponding experimental Windows build to create. To find
here Win64Bit and www.graphicall.org/builds/builds/showbuild.php?action = show & id = 1525 (here as Win32Bit). 