It's hard to say, because there are at least three different facets that we know little about.
One is the additional eye candy and DirectX 11 support in TW3. Both of these are expensive, especially if DirectX 11 is not used conservatively. This could mean more demanding graphics if the game implements all of the promised features; it could just as well be less demanding if their use of DirectX 11 is very careful.
Another is the optimization in RedEngine 3. They surely did not start from scratch and write a whole new engine. It will have large pieces of the Witcher 2 engine, and these may be optimized to varying degrees.
The third, and to me the most important, is we do not know what the limiting resources are. Optimization is nothing but a waste of a lot of time and money if you optimize the wrong code or the wrong feature. You have to find the limiting resources and optimize for those. In the Witcher 2, the output processors (ROPs) were very noticeably limiting -- to the point that you can make predictions of performance within ten percent or so from knowing the ROP complement and the core clock rate alone. In Red Engine 3, well, we just don't know yet. They may have put more burden on the shaders and less on pixel-level processing at the output stage.
So any prediction of whether the Witcher 3 will perform better or less well on the same hardware is really a wild-arsed guess.