Is overclocking good way to save on energy? (good article)

[nyker96]

Is overclocking good way to save on energy? Check this out:

http://www.xbitlabs.com/articles/cpu/display/power-consumption-overclocking.html

 

[ViRGE]

There's nothing particularly surprising here. It's voltage that makes power consumption really spike; CPU speed increases at stock voltage are neutral if not slightly more efficient.

However both should pale in comparison to driving down the voltage.

 

[deimos3428]

Not surprising. You really want to be on the very edge of stability at all times -- just not over the edge, and that's the tricky bit.

 

[nyker96]

looks like most processors 45nm are OCable to about 3,5-3,6 at default volts that's the sweet spot for energy saving. the 32nm parts seems to go higher. still I thought it's interesting to see some hard data to support this 'know secret'.

 

[aigomorla]

Problem with articles that tell you set OC is..

You can never write an article and say this is the typical OC, nor a typical power draw.

Because each chip you get is a YMMV to the MAX.

You may get a super cherry, you may get a back seat granny.... its ALL YMMV!
Your chip can run super cool, your chip can run super hot... it could be super conservative with voltage, it can be HELLA leaky with voltage...

The only thing intel promises, is that it will run @ speed @ default voltage for 3 yrs.
Thats it...

 

[MagickMan]

I know my system pulls quite a bit more current when I run @4.2GHz, which is why I usually stay at 3.8 for most things (my max clock at VID), unless I'm benchmarking or doing some intense encoding.

 

[solofly]

1000Mhz overclock costs you around 100 watts...

 

[EarthwormJim]

There's nothing particularly surprising here. It's voltage that makes power consumption really spike; CPU speed increases at stock voltage are neutral if not slightly more efficient.

However both should pale in comparison to driving down the voltage.

Wouldn't switching losses increase proportionally to clock speed increases? Are switching losses just that insignificant?

 

[Hacp]

Isn't power determined by c*frequency*voltage^2? I saw that equation somewhere.

 

[fffblackmage]

Wouldn't switching losses increase proportionally to clock speed increases? Are switching losses just that insignificant?

Isn't power determined by c*frequency*voltage^2? I saw that equation somewhere.

Yup, switching losses are there, but it's just not as significant as raising the voltage. You can actually see very noticeable changes in power consumption when overclocking and increasing the voltage (using Kill-A-Watt meter or something).