Socket F (LGA1207) will supposedly support FB-dimms and up to 32-way (32 sockets) configurations, so the pin count cannot be compared to socket AM2's directly. Hell, it'll probably have an increased amount of HT lines in order to make 32-way possible without huge performance drops.
Having a base HTT of 333 means nothing since I doubt the multiplier will remain at 5x (a 3x multiplier would leave it at exactly 1000MHz). Even if it does remain at 5x I doubt there will be any benefit from having a higher Hypertransport speed, since A64s are not even close to HT contrained (MP Opterons are, however).
Zebo, I truly don't see much of a benefit from having an integrated PCI-E bus unless a video card that need lots of memory accesses (like a craptastic turbo-cache one) is attached to it. If there was a sizable performance increase from dropping latency to the southbridge devices (like audio, drive controllers, network adapters, etc) then all of these things would have been thrown into the northbridge long ago. Even video cards don't seem to benefit much from having lower latency, since ULI's southbridge AGP 8x (on the 1695/1567) performs as well as some of the best northbridge-based AGP chipsets. If there IS a performance increase then great, but I dont see how it could happen. I truly hope that AMD doesn't start adding crap that does not contribute to performance simply to claim that it has more "features", especially considering that anything thrown into the CPU needs pins on the package.