Technical bla bla:
Hyperthreading in its initial P4 era form was basically a second set of a the basic x86 registers (pointers and data for basic ops) in an effort to be able to slide a bit more into the pipeline (and counteract the P4's very low IPC, or "instructions per clock" cycle.) Although Intel would claim 30% improvement even in the early P4 era, this just wasn't true for every situation. However for users of software like MS Office (which is largely pushing stuff around in memory, ie, using basic pointers) the net effect was a small benefit in performance but still only 30% in specific benchmarks and not actually across the board (gaming is well known for designing new software to meet specific benchmarks as well.) Hence those were the benchmarks quoted in this era. Steinberg also claimed similar performance benefits in this era, but I was under the first 2 Nuendo licenses during this time and tbh never saw that with HT enabled (even after I moved Scope to a 2nd PC.) Leaving HT disabled and having a proper dual CPU machine brought far more gains, but even in that case it was typically only 30-50% more 'dsp' power for audio applications that supported multiple cpus. There are some differences between multi cpu and multicore (in regards to having data local to caches--contention--and memory latency issues) but overall this is much closer to what multicore cpu's are today (ie, more than 1 'complete' logical processor.)
The problem back then for users who are relying on their cpu for DSP operations or heavy number crunching was that the OS had a tendency to try to make use of the extra 'virtual' HT core to schedule additional work that was unrelated to your audio tasks (basic background services etc) and (in many cases) that reduced the performance of the audio operations you were interested in because the OS was treating your workload the same as it would any average office user at the expense of the DAW application. Setting things to background scheduling actually made this a bit worse, something that is still common now and was necessary in the single core era to keep the ASIO buffer handling happy (delivering the buffer to the card in a timely fashion under high cpu load.) Intel made improvements in the Hyperthreading implementation over time as did OSes (win/linux) and applications until certain benchmarks showed improvements even for audio users (though I personally kept it disabled especially with Scope & software running alongside.)
The modern implementation of HT is much more robust and mature, and basically allows for the scheduling of a far more complete
cpu state than the P4 cpu's did, so the benefit of having it enabled is now far greater, even though modern nehelem based cpu's don't have the IPC issues that the p4 era cpu's did.
Modern Multicore cpu's with the i7 brand name (both Lynnfield LGA-1156 i7's and the "full" LGA-1366 i7 versions) also have hyperthreading enabled, while the i5's (and upcoming i3 etc) do not (though it's still present on the core, just disabled by burning the traces as with celerons and other products with 'cut down' features to be able to 'save' some poorly binned products which would otherwise have lower yield.)
---------------------
Also modern i7 branded cpu's will have different levels of "
TurboBoost". To understand what this means you need to track power saving features back to the introduction
Speedstep and of early clock/sleep/power states.
In more detail: the early P3 the cpu multiplier would switch between its lower and upper settings (Speed"step" between 2 states) and mobile cpu's could also additionally control voltage (Enhanced SpeedStep).
The next implementations (P4 mobile) had only a few speed increments, for example a 1.7 GHz Pentium 4M would clock down to 1.6 GHz, 1.2 GHz, and786 MHz (lowest sleep state besides suspend.)
The next iteration after that (full p3 derived Pentium M used under the Centrino brand) introduced "EIST" which could be "stepped" over the range from 600 Mhz to 1.6 GHz (200 Mhz at a time) while adjusting the voltage from 1.484 V to 0.956 V in addition to controlling the cache as well (which is external to the core and runs at a different speed.)
Since the Core1 & Core2 were derived from the Pentium-M, this became "Enhanced EIST" where even desktop cores could have relatively fine-grained control over power states, giving us multiple "C" states and "P" states as a result (clock/voltage.) By Core2 each core could be controlled independantly as well as the independant cache & shared L2 cache (Enhanced EIST is defined as that which is present on multicore cpu's with a shared L2 cache.)
Something not mentioned yet, but the time window for these 'power saving' states also get smaller and smaller with each iteration, up to the point where the modern "Enhanced EIST" has very fine-grained ability to clock cpu parts up & down as needed to make efficiency higher.
Which brings us to the present Nehelem family i5 & i7's, where the cores can not only be individually clocked down, but also UP (Turbo) and additionally completely turned OFF (something which wasn't possible before, as even the lowest sleep states still had the cpu cores & caches drawing power.) So this clocking UP is known as "TurboBoost", and takes advantage of the fact that the cpu is designed to work within a certain "thermal envelope" that equates to heat output based on power draw. When all cores are not needed for a given workflow, the energy & heat that is 'saved' from disabling 1 or more cores is 'used' to clock up the cores in use to provide a "TurboBoost" to the tasks being handled. The idea is that applications which aren't optimized to make best use of all cores present will still get a benefit of a modern cpu (as otherwise nehelem isn't always a win compared to Core2 with applications which don't rely heavily on memory bandwidth.)
------
Incidentally there is also some evidence that the increased interrupts caused by E-EIST and turbo modes can degrade certain realtime operations, you can limit the power saving features by choosing your Power modes in Windows (and in Vista/Win7 you have the ability to dig in a bit deeper without relying on additional tools as was necessary in Xp to tweak specific profile settings.)
I do think that we're on the cusp of the market fully leaving Core2 though and there's no way that Apple can ignore that. It means leaving behind nvidia as the chipset, though it's hard to say who they will choose for the gpu (certainly not Intel's integrated gpu for the upper tiers of the portable line.) That actually brings up another point though...