I suspect (but don't really know) that there may be some benefit in certain applications -- but it would be very difficult to quantify.
You would simply have fewer cores sharing the L3, and this would be a boost. You might want to check something like single-thread CineMarks between PhII X2s & X4s at the same clocks.
What you essentially would be showing is that even cores at rest may have flushed enough of something into the L3 to ding performance slightly -- but it's a little misleading, as the AMD arch seems to get diminishing returns from piling on the L3.
In other words -- adding 10% more L3 gets you a 5% boost in performance; and adding 20% more L3 might get you a 7.5% increase in performance.