Moose,
It's true that there will not be a linear speedup. In some cases it may be more efficient to render multiple instances and limit the number of threads per instance. However, some advantage can be gained by rendering the image in one instance on multiple cores, and that is from the decreased RAM use.
You will be able to limit the number of threads that TG2 renders with, so it will be possible to experiment with different rendering strategies.
I don't want to release any benchmarks at this stage, but early testing shows speedups on 2 cores and on 4 cores that are very reasonable for some scenes. It remains to be seen how well this applies to a wide range of scenes, especially more complex ones, so it would be foolish to give any figures just yet.
Matt