You wanted proof, here it is:
http://www.hypertransport.org/tech/index.cfm
http://www.hypertransport.org/images/features_chart.gif
HT1 has a max BW of 12.8GB/s at 800MHz operation.
Considering that we are talking ONE SOCKET SYSTEMS (Phenom, not Barcelona), that means even AFTER you subtract out the 6.4GB/s for full utilization of a dual-channel DDR2 controller on the memory bus, there is a full 6.4GB/s left over.
All CORE-TO-CORE COMMUNICATION on X2 and Phenom CPUs occurs via the crossbar or L3 cache, respectively. So there is STILL nothing that can fully utilize the extra 6.4GB/s BW in HT1, since CORE-TO-CORE COMMUNICATION has a dedicated pathway.
EDIT - HT3 on the desktop is like going from a 4 lane to a 16 lane highway when you are the ONLY car on the road in both cases. Just because you have more lanes, it doesn't magically make your car faster when there is nothing to impede it in the first place.