First off, HTT is a VERY weak form of SMT; only about 20% performance gains, but at a very low die cost (EG: Cheap to implement).
The main problem is that while it is easy to make reasonably independent threads (games generally have 50+ going at one time), only a handful (3-4) are going to do any meaningful amount of work. As a result, 2-3 cores end up doing the vast amount of processing, while the others handle the lower-volume threads. Hence why getting more then a quad is a questionable purchasing decision.