I really like Blender’s compositor and VSE, but when doing some heavy compositing it takes a while to render each frame, I have a quad core 2.4 GHz. Turning on Opencl doesn’t seem to make it any faster. I have a NVIDIA GT 240, would upgrading to a GTX 580 make compositing faster?
OpenCL doesn’t do a whole lot except for a couple of nodes (like defocus), unless things have changed recently. A few tricks:
-I think this might go without saying, but don’t use the renderlayer node in any final shots, it requires the input layer to be re-rendered everytime. Render once, save as multilayer EXR, and comp with that.
-Render from the command line, there’s a very noticeable delay to put the final composite up in the GUI. GUI renders have trouble doing much better than 1sec/frame. Simple comps will run WAYY faster when done headless.
-Try lowering the chunk size, it can help a bit on lower resolutions/lots of cores. Experiment, too low can make things worse.
-This might not apply to you, but if you are using a render manager and have any boxes with 12-16+ threads, try setting it to run several instances at once on different sets of frames. I’ve noticed the compositor has trouble finding work for more than 6-8 threads at a time. (at least on the 720p batches I do at work a lot).