"Even if the gap is only a few seconds, the CPU temperature has already dropped from the highest points reached during the test."
Shouldn't that be the GPU since 3dmark is what is referred to?
Yes, you're right. Typo.
What's your impression, does the thermal blocks for the motherboard make any big difference?
I'm not prepared to say BIG
difference, but definitely, those things have
to help. They're the right height to fit pretty tightly, and there's no question that thermal conduction via the underside of the board does have a real effect on operating temperature. The fact that the EndPCNoise TNN500AF system was cooling the board the same way with a Prescott at ~100W suggests to me that this cooling strategy is pretty effective. I know Zalman got grief over VRM failures in their first iteration of the TNN case, so they had to come up with something that works. And there is some convection airflow over the long term -- you can sense it at the holes around the top after the system has been workling hard for a while.
I'm still surprised that there wasn't more of an attempt at drive silencing by Zalman
I'm not. You have to remember that even in the demanding "silent sector" SPCR forum activists are more extreme and demanding than most. Lots of people feel anything less than ~30 dBA@1m is "silent", especially if it's not a pure tone. Most people readily put up with a lot more noise that I do -- or you, it sounds like. They just shrug and think, "oh well, that's the way it is...
" This is the reality that most PC companies consider when developing products. In my consulting work, usually the biggest challenge is to justify challenging low noise targets even to clients who come to work with me on low noise products.