Nope, AMD seem to be sticking with the Bulldozer-like architecture for now (2 modules each containing 2 integer and 1 floating point unit), as they're hoping parallel/floating point workloads will be sent to the GPU portion (where 80% of the power is) as soon as the software has been updated. This is part of the reason why it looks like an i3 in floating point tests.
Ah, so that explains why kaveri's cpu performance isn't matching AMD's FX CPUs in some tests as the FX still have a clock speed advantage.
I guess we really need to wait for software to catch up at this point.