Intel really need to re-evaluated their efforts in SMT, HT is and has been pointless for a long time now.
Complete bunk. HT is a fantastic way extracting an extra 30% of performance out of existing silicon by interleaving instructions from 2 different streams during waits from memory accesses, cache misses, branch predictor miss, etc. It obviously won't benefit anything that doesn't run more than 4 (on an i7) strong threads, but for applications that do, the performance gain is impressive, especially considering how little silicon HT takes. HT (and the L3 cache) is the reason that the difference between the i5-2500K and the i7-2600K (only 100 MHz clock speed) is much greater than the difference between the i7-2600K and the i7-2700K (also just 100 MHz, but no difference in HT support here).