I only want to say that a lot of this trying to save power and cut down on the power when in idle etc is generally not worth the effort. The amount of power actually saved is negligable and while the idea is admirable the reallity of it is fairly futile.
I'm surprised you should say this. If I offered you two CPUs of the same make/model one which overclocks to 3.8GHz @ 1.4v and the other which overclocks to 3.8Hz at 1.5V, which would you choose?
So why would I choose to operate a CPU at 800Mhz with a Vcore of 1.325V, when it can work at 0.8v, or operate it at 3Ghz with a Vcore of 1.325v when it can work at 1.125V?
I don't have the means to measure the CPU power draw on the 12v line. But there are some published results here:
http://www.lostcircuits.com/mambo//index.php?option=com_content&task=view&id=103&Itemid=1&limit=1&limitstart=6The idle Power draw for the 960T is likely to be around 10w, but that's only if Vcore is around 0.8v. While a CPU is not a simple ohm's law device, the power draw is still reckoned to be roughly proportional to the square of the Vcore. Operating the 960T at 800Mhz with a Vcore of 1.325V would consume approx 25W. Likewise, operating at at 3Ghz with a Vcore of 1.125V instead of 1.325V may use around 20% less power for the CPU.
Of course, the time a system spends in a low CPU state depends on its use, but I believe it's easy to underestimate this.
One of thses days I'll get myself a power meter and attempt to measure at least the system power draw and get some actual figures for the mix of power states my PC uses. At current UK electricity prices, based on my PC up time, I'd save about £1 per annum for every watt saved. It could amount to £30 per annum. Not a lot, but not futile, nor does it require much effort to have a 960T operating a more sensible voltages.
So until I can show otherwise, I prefer to lower Vcore for all operating frequencies, reduce power draw and produce less heat. The system is cooler, fans can be ramped down and noise reduced and the strain on the m/board vrm is less.