First of all undervolting is a lot like overclocking - you don't do it without testing stability using something like Prime. Secondly, when idle, other factors like C3/C6/C7 states have much more of an impact (they completely gate down parts of the CPU), so changing frequency and even voltage doesn't make that much difference below a certain level. It's under load that most of the difference are seen. Eg, a -0.1v undervolt on my i5-3570 (down to 0.696v idle) knocks off barely -1w idle (and there's 0w difference in 800MHz vs 1.6GHz idle @ same voltage), but -10w load at stock 3.4GHz, & around -18w load with a mild OC to 4.2Ghz. Temps fall by between -5 to -8c on a 212 EVO depending on load. Load fan speeds reduced by 200rpm, etc, so it's definitely worth doing for HTPC's / silent PC enthusiasts, etc, to gain ultra-low 600rpm silent fan speeds that never ramp up under load.Have anyone of you undervolters tested what are the energy savings by undervolting? Like I said before, I undervolted my 4130 processor to 0.75v and I checked with my energy meter after that how much lower energy consumption I got from that and the end result was none. I mean zero, I did not get even one lousy watt less in idle consumption.
So in my case by undervolting there was nothing to be gained, but at the same time you risk your system to become unstable.
Yes, there's a slight risk of instability (as with overclocking) but then that's what stability testing is for. It won't make your PC draw less power idle (a platinum vs bronze PSU or LED vs CCFL LCD monitor is more likely to do that), but it can knock off up to 20w power & up to 20c temps under load depending on CPU & cooling system, ie, a compact mini-ITX case with limited small slimline cooler will benefit more from it in terms of reduced temps than a full ATX with water-cooling.
