Intel Hyperthreading shares internal resources of one core between 2 threads. I do not recall, but it might be call hyperthreading not hyperprocessing because the VM translation is shared, so they must be 2 threads of the same process (also less security worry that way). Google the wiki. AnandTech does in-depth CPU analysis and benchmarking.

This is a pretty wide ranging subject. One new chip lowers power and speed on idle cores and raises on remaining busy cores, Turbo Boost as I recall. Otherwise, all things being equal, CPUs on different RAM/motherboard (clustered) run faster than on same RAM/motherboard (SMP), which in turn run faster than on same chip, l2 cache and FSB, which in turn run faster than when hyperthread-sharing the CPU. A good OS CPU dispatcher may be able to exploit all SMP CPU speed tradeoffs. Some architectures flush various lavels of cache, registers, instruction buffers and VM translation as you change threads or especially if you change processes for both accuracy and security, but others benefit by keeping one active job as the only one on one CPU, where some or all that stuff is already loaded with good data more local than RAM.