I have compared the performances of a GTX Titan and GTX 780 using the same machine under win7 64bits with 320.49. According to the specifications of each graphic card (cuda cores: 2688 (Titan) / 2304 (780)), we can make the assumption that the Titan is faster than the 780.
However, I have made common benchmark tests (FurMark, TessMark, 3DMark, …) and, except for tesselation, the GTX 780 is better in all tests. In the same way, with a lot of other 3D apps (e.g. http://blog.icare3d.org/2010/07/opengl-40-abuffer-v20-linked-lists-of.html ), the GTX 780 is faster with a factor range between x2 and x10.
I have tried different driver versions and the results on test  changes from 60fps (320.49) to less than 10fps (327.23). Note that the same test with the GTX 780 runs at 120fps (320.49).
So, first, I’d like to know if I am wrong when I think that GTX Titan is faster than GTX 780 for rendering operations ? (Of course, for computing it may be not the same).
Is there a problem with the WHQL drivers since the more recent one seems to be less efficient than the previous one ?
Thanks for your help,