[germanium]

I recently built a Core I7 setup for myself. I was impressed with the overclockability but was not really satisfied with the amout of fan noise it was generating though I did love the performance. So I lowered the clock speed back down to stock. This didn't seem to improve things that much by itself so I started lowering the voltage below stock voltage. I reached .950 volt setting with good stability. Stock the voltage was 1.20 but the motherboard if left on auto automatically overvolted to 1.225 volts. @.950 volts the processor draws about 40% less power than stock making the Core I7 an extremely power efficient processor indeed.

This voltage was achieved without turning off any of the processor features as some claimed was nessessary to get even to 1.000 volts!!This is probably more efficient than my former Core 2 Quad Q6600 proccessor was & likely more efficient than even the latest 45nm Core 2 Quads.

This made a huge difference in the fan noise. The biggest culprit being the power supply fan which now barely even ramps up under full load whereas it roared even at 50% load before.

 

[MCC]

I'm running the Core2 Duo CPU in my Thinkpad @ 0.8750v @ 800MHz all the way to 0.9500v at the full 2.4GHz. I'm willing to bet that the CPU could go even lower but I've hit the lowest setting that the hardware will allow. It's 100% linpack stable for 20 iterations on each multiplier.

End result: the fan doesn't even spin most of the time. This is significantly down from 1.0-1.250v stock.

 

[germanium]

Nice to see the 45nm Core 2 Duos can go that low too. At stock speed my minimum speed is still double the speed of that Core 2 Duo though the limit seem to be about the same at the highest stock speed. the .950 setting results in a real voltage under even idle load is .928 volts running at 2.66 GHz. I found no power savings running the lower multipliers though at this low voltage it refuses to turbo up to 2.8 GHz. A good thing as it not stable there anyway at least not at this low voltage.

Wonder how the 45nm Core 2 Quads do? I suspect they would need more voltage to reach the same speeds, I could be wrong though.

 

[MCC]

I'd guess that 45nm C2Q chips capable of 0.95v would be rarer since there's a larger surface area and therefore a greater chance of impurities. I'm not as concerned with efficiency on my desktop (SLACR Q6600) so I let it run @ 1.4v/3.3GHz/1.9GHz FSB 24x7. I had attempted to use throttling in the past but I discovered that it would throttle down at completely inappropriate times, i.e. when FFDshow can't keep up and my video's already lagging waaayy behind. ISU covers the power bill so no problems there.

Edit: And yes, a 3.3GHz Core 2 Quad can't keep up with realtime video decoding esp. when there are fancy subtitles onscreen. A Core i7 would be awfully nice.

 

[germanium]

Did Some further tests to get the real scope of power savings undervolted at stock speed compared to stock voltage but overclocked to 3.6 GHz & the power saving is very impressive at full load.

Overclocked system @ 1.20 volt setting
Idle: 162 watts
full load: 260 watts

Undervolted .950volts @ stock speed 2.66GHz
Idle: 135 watts
Full load: 180 watts

As you can see overclocked @ stock voltage the power went up almost 100 watts under full load but only went up 45 watts under full load undervolted @ stock speed. To get a better Idea how much saving there is I will try stock speed stock voltage next.

Stock voltage @ stock speed
Idle: 153 watts
Full load: 234 watts

Here the Idle wattage is less than 10 watts below the overclocked wattage at Idle & the full load watage jumps 80+ watts compared to only 45watt jump when undervolted

 

[Zodduska]

What are you using to load the CPU and test stability?

If you are not using it already I suggest a multithreaded 64bit prime95 blend test for 10+ hours
DL here:
http://files.extremeoverclocking.com...ef83366a55de6c

 

[Iron_Dreamer]

Generally speaking, if a chip is highly overclockable, it can also be ran at much lower than stock voltage at stock speed. Or a combination thereof. I can still undervolt my E8400 by .1v from stock, even at 3.6GHz. Recent Intel chips are generally very good in this regard.