Script updated, added Urs' CPU AP detection, x32 CUDA MB detection and columns rearranged for easy Copy&Paste into spreadsheet.
ATI GPU is HD4870, NV GPU is 9600GSO.While NV outperforms ATI on mid and high ARs, this ATI GPU clearly wins by elapsed time (CPU consumption still bigger) on VLARs.
Script and example of its result in attached file.