Hmm. I don't know if I would use the ease of seperation as a measure of the thermal efficiency of TIM, I've seen HSF/CPU pairs using phase-change pads, that were a PITA to remove, but the bond between them, in terms of distance, wasn't all that close. I've also had to remove heatsinks from ancient CPUs attached using generic paste, and it had long since solidified and dried out and slightly browned. It was very nearly cemented on. But again, it wasn't making a very close contact. The ideal goal is to fill in the microscopic crevices that do not make absolute physical contact, while at the same time, NOT increasing the distance between the two surfaces that does make contact.
So in that light, I actually would rate generic paste (properly applied) over phase-change pads, in terms of cooling efficiency, and I suppose AS, since everyone seems to be a fanboy about it, although I don't think that it is all that better than regular (good-quality) paste, when applied properly (no more than 1-2C diff).
But in terms of long-term maintenance, the phase-change pad is probably the best, needing, AFAIK, no real maintenance, whereas generic paste may need to be replaced every two years or so, and AS, I don't really know, probably not as often, say 3+ years.
Another thing to consider is that those TIM pads aren't all the same. Some are spongy, and some actually have a tiny metal shim that is covered in TIM material on both sides. Of course, it generally is better to use one without, and paste or AS would definately give better thermal conductivity, since there is only one TIM layer total, instead of two. The only consideration there is whether the tiny metal shim is needed for mechanical reasons.
Btw, as far as non-AS pastes go - there is a difference in terms of quality levels. Not all generic paste is the same. Perhaps that's why some swear by AS over paste? I've used mostly only higher-quality stuff, perhaps that's why I haven't seen much difference to AS in terms of thermal performance.
