Please help us understand the Cool and Quiet. I thought that this feature would slow down the clock rate if the processor started heating up due to heavy load.

This gets confusing. Before we got into the throttling with the agent lets just talk about the benchmark. The benchmark is run at the end of a work unit. The benchmark, like the science program is primarily floating point so it stands to reason that the benchmark will likely run at about the same speed as the science work unit. At least that is the theory. But the theory was developed in a time where processors ran at one speed. So the thing made great sense. Now that we have things like Cool and Quiet and a default throttling by the UD agent, the whole thing gets quite complicated.

You can look at:
http://www.unofficialworldcommunitygrid.com/Overall.shtml
and
http://www.unofficialworldcommunitygrid.com/benchmark.html

When complicated by the issue of people trying to "cheat" it gets more convoluted. It seems to me that it will take multiple benchmarks and modifications to help with the accuracy (which is done in BOINC) and perhaps other changes. But then, there are still ways to "cheat". Very messy business. I think I understand your concern but I wonder if it is all worth it.

Kind Regards to all

Nope, that's overheat protection throttling. A64s and Pentiums do that too, but it's completely different than Cool 'n' Quiet.

C&Q is the same idea as Intel's SpeedStep. They both slow down the CPU when it's idle, so that it runs Cooler (and Quieter, assuming your cooling system is temperature-sensitive) when you don't need full power. When the CPU gets busy again, it ramps the clock rate and voltage back up to provide full power when needed.

It works really well, in the normal PC usage scenario where it's idle most of the time, with sporadic bursts of high usage. My A64 is set up so that when idle it runs at 960 MHz and 1.1 volts, which allows SpeedFan to turn the CPU fan completely off and still have it stay at only 38C. At full load, it ramps up to 2.64 GHz and 1.45 volts, providing (somehwat overclocked) power on demand.

The overheat protection throttling kicks in if, at full load, your heatsink isn't able to keep up and the CPU starts to overheat. If this happens, the CPU will reduce its clock rate back down somewhat to avoid damage. This shouldn't happen though with an adequate cooling system.

If your PC is used for crunching 24/7, C&Q won't really matter, since your PC will always be at full load and full power. But it shouldn't hurt anything, either. It's very responsive, and adjusts the CPU speed several times per second according to the load at that moment. It shouldn't normally affect the benchmark results much, if at all.

The CPU would be at full speed while crunching the WU. When the WU completed, the CPU might drop back to slower speed while the results are written to disk, since the CPU is mostly idle during heavy disk I/O. But within a half second or so of the benchmark starting up, it should be back at full speed and giving pretty much the same benchmark result as if C&Q were disabled.

Hmm... unless the UD benchmark is really short? If it's more like a clock rate estimator loop than a real benchmark, it may not run long enough to fully ramp up the CPU speed, or to average out the initial low speed period. But if that were true, I'd expect a lot more users asking why their UD CPU rating was so low.