We already know how much better Haswell Xeons are than the Ivy Bridge Xeons in the current nMP. We also already have a decent idea how much faster Broadwell architecture will be relative to Haswell, even though Xeon Broadwell-E5s are not available. More or less you should expect about 10% clock-for-clock performance increase from Ivy Bridge to Broadwell. But that’s based on the desktop/mobile core series which had more emphasis on power consumption and iGPUs. So the Xeons may be a bit better than 10%. But what we don’t know is how Intel will juggle the clock count vs clock frequency across the line. Its probably safe to assume they will still have 4-core zipping along at near 4.0 GHz though. I just don’t see them dropping the 4-core as the base E5-1600 CPU yet. But it might be possible that the middle ground 1600 is a 6 core (i.e. replacing the 1630 4-core). Then, I would guess they also get an 8-core down in a more modest price range (i.e. the 1650 becomes an 8-core), while the if have you ask you can’t afford it version 1680 gets a 10 core, maybe even a 12 core.
This is all conjecture, but Intel has been slowly adding extra cores at the lower price points, while taking advantage of turbo boost to not loose out on clock rate per working core, for several generations now and I doubt it stops here. Before we know what intel does though, we can’t really figure out what Apple will do. And until we actually see tests of those Xeons we won’t know for sure, but my guess is you’ll see 10-15% increased performance per core, but also 2 additional cores for the same price (except at the base model), which MIGHT mean about 40-50% increased performance in applications that can put all the cores to work [i.e. 8(cores)*1.1(speed increase per core)/6(old core count) = 1.4667].