# Overclocking & Power

Tags:
Last response: in Components
Share

When overclocking a CPU by raising the FSB, is there a direct ratio of power draw?

For example, if I increase the FSB by 12%, does the CPU draw 12% more power?

What's more, what percentage increases do the other parts rise by?

In short, I have a decent 300W supply in one very stable PC, and plan to oc it by 12%. I reckon an Antec true330W will do the trick, but I wanna know if it's a must or not.

<b><font color=blue>~ <A HREF="http://forums.btvillarin.com/index.php?act=ST&f=41&t=32..." target="_new">My System Specs</A> ~<font color=blue></b> :wink:

Overclocking Expert
CPUs Expert

First of all let me ignore watts as a unit of heat (uing joules instead) and use it only for a unit of electrical energy.

Now to your question. It's not linear because heat affects resistance. A hot processor could in theory draw more power than a cold one. The published numbers for heat output are theoretical, based on testing I think, most processors of a certain clock speed having roughly the same heat output. But when you overclock you often push a CPU past it's tested limits, you need to raise voltage, you get more heat.

Take the PIII for example. I had a 700@933. It put out a lot of heat. Now, I had a 1000EB at 1000, it put out very little heat. Both used the same core design, one was simply better than the other. At 933 and 1000, both had the same bus speed, but the faster one ran cooler.

Why go into so many details? Because I know that there will be people to argue against me on this. It doesn't go with what they've learned about CPU's in their classes. But classes don't teach you everything. We know that you can convert heat from joules to watts. And the more joules of heat, the more watts of power it takes to make them.

So I've already shown that an overclocked CPU can draw more power than a stock CPU based on the same core and without speed being the factor, since the cooler CPU was faster.

The two major contributors to heat of course are clock speed and voltage. The more clock speed, the more pulses going through, each pulse creating a certain amount of heat. To those who want to argue against this, think of your typical incandescent light bulb...if you turn it on 1 second per minute, it will never get really hot, but if you turn it on 60 times that, it's on continuously and gets very hot. 30 1 second intervals in a minute would produce less heat than 60, but more heat than 1. Processors are NOT resistors, don't argue Ohms law exclusively, as a procesor unlike a resistor has not a continuous flow.

Ok so the other thing is voltage. Again I'll use a PIII as an example. If it takes 10% more voltage to make a PIII 750EB run stably at 1GHz than it does a 1GHz PIII run at 1GHz, the 750@1GHz should be 10% hotter. And this DOES follow Ohms law without taking to many other things into account, because V=IR, so V/R=I, as V increases I increases with R constant. I is in Amps, volts times amps = watts, watts convert back to joules...

OK, so you could probably work out the heat due to increased voltage fairly easily, but what about the heat due to clock rate? It's probably a curve, know any calculous? LOL! Using the lightbulb as an example, the heat barely increases when you increase from 1sec/min to 1se/30sec. But then it probabably climes a steep slope in the middle, and then tapers off to 1 constant high at 60seconds per minute. Room temp would be your lower limit, full power temp would be your upper limit. Heck, when I think of it that way, my Calc classes and my CPU Overclocking limits both make sense! OMG, I had an epiphany!

<font color=blue>There are no stupid questions, only stupid people doling out faulty information based upon rumors, myths, and poor logic!</font color=blue>
Overclocking Expert
CPUs Expert

My answer is NO, lol, or at least not exactly. At some points in the curve it would <i>appear</i> to be a straight line, at the end of the curve it wouldn't. So <i>almost</i> yes at speeds near stock, and definately no at extreme overclocks would be my best estimate.

<font color=blue>There are no stupid questions, only stupid people doling out faulty information based upon rumors, myths, and poor logic!</font color=blue>
Related ressources

Raising the FSB doesnt lead to greater power draw (for the cpu) provided the Mhz of the cpu remains the same by lowering the multiplier.

If the CPU mhz goes up then so too does the power draw.

And some CPU's just draw less power than others... even tough they might both be p3's or palomino's. It depends on the "goodness" or the core and core revision you have.

I think your antec true 330W should be sufficient... just watch the 5v and 12v rails for excessive dipping under full load.

<b>My Computer is so powerful Sauron Desires it and mortal men Covet it, <i>My Precioussssssss</i></b>

PIII so multiplyers don't count. :smile:

I realise they will draw extra power, but I wanted to know if the extra power draw was in direct proportion to the percentage of overclock.

<b><font color=blue>~ <A HREF="http://forums.btvillarin.com/index.php?act=ST&f=41&t=32..." target="_new">My System Specs</A> ~<font color=blue></b> :wink:

i believe so yes... a quasi linear relationship, or at least close enough to linear for the small range of overclocks that most obtain.

Voltage however increases power usage in a square relationship though.

<b>My Computer is so powerful Sauron Desires it and mortal men Covet it, <i>My Precioussssssss</i></b>