I have a theory about why you might see this improvement in single core performance when benchmarking. This is completly hypothetical so feel free to correct me.
So with AMD's Bulldozer architecture (including updated versions like Piledriver used in the FX-6300), AMD uses shared resources between pairs of cores. By doing this, they can essentially create two cores where there would traditionally be only one. This should also help with multi-threaded programs because it can process more information quicker. However, when bench-marking, this could saturate the CPU core's and mean each core has less resources to work with than it would if only one core was enabled. This could explain why the single core results are increased when most are disabled.
You would realistically not notice any improvement because even if a program is only single threaded, it would only use one core and not affect the other core it shares resources with. It would be detrimental if the program/s is able to use all of the cores however which would explain the lower results overall. Either way, I think it would be better to leave all cores enabled for real-world applications (like gaming).
Just as a test, what happens if you disable core's 2,4 & 6? Does the single thread performance increase?
Further reading:
- http://www.futurelooks.com/the-amd-fx-8150-bulldozer-cpu-and-scorpius-fx-platform-reviewed-part-one/