It isn't a fair test because TG3 renders to a higher detail with 4 times the number of micropolygons. Having said that, it is reasonable to expect that with 4 cores (or 8 virtual cores) TG3 @ detail 160 would complete at least as quickly as TG2 @ detail 80, but in the default scene it doesn't. There are a couple of reasons it doesn't. First, the subdivision and displacement phase of the rendering is not multithreaded yet, only the shaders and atmosphere (including lighting) are. The default scene is very simple and a disproportionate amount of time is spent in the subdivision and displacement phase. Scenes with more demanding lighting, atmosphere and shaders will typically preview much faster in TG3 than they did in TG2, although, again, the higher maximum detail might seem to nullify that difference if you are waiting for the preview to finish all the way to 160. If you are happy with detail 80, you can pause it when it says "rendering detail 160". The second reason TG3 isn't as fast as you might think it should be is because the thread scheduling algorithm for the 3D Preview needs to some improvements. You can expect it to get faster in future. The preview in 3.0 should almost always be faster than 2.5 though, for equivalent detail.
Matt