Is CPU power draw/wattage linear with CPU usage?

S

spotpuff

Gawd
Joined
Aug 15, 2001
Messages
522
I'm just wondering if, for example, a CPU draws 50W at idle and 100W at full load, of it's at 50% load will it draw 75W? Or will it be some fancy number in between 50W and 100W?

A lot of reviews compare power at full draw and idle, but I would imagine that typical computer usage would be somewhere in between with programs using a little cpu power here and there, though that could be wrong.

Interested because TR just posted a review to a low-power athlon, and I'm wondering how the mobile athlons fare vs the c2 duo mobile chips.
 
My system is connected to a CyberPower battery backup, which comes with monitoring software that tells me just how much power the computer (including dell monitor) is drawing. The list of other parts is in my sig.

At idle, it draws 202W. Under Orthos load (roughly 100% usage on both cores), it draws 222W.
 
20W swing from full to idle seems low, but I was more interested in moderate usage, say 50% CPU usage, how does that power draw compare?

20W is low compared to what I've seen on Tech Report's reviews of CPUs at idle vs load; typically the swing is 50W+ and sometimes much higher.
 
you have to remember what watts are what.
watts= power draw (amps) x current (volts)
if your video card uses 100 watts and runs at 2 volts, then its using 50 amps.

but now your household power is 120v. (or 240 for you damn foreigners, but say 120 for simplicity) if you use 20 watts at 120 volts, thats a little less then 0.2 amps. that 222watts your reading is your household current, 222watts is around 2amps at 120v. but at 12 volts (after the power supply) that would be around 19 amps.

if any of that made sense, it illustrates why a processor rated at 100watts from the power supply does not mean 100 watts from the plug in the wall.
 
Fair enough, make sense.

Yeah the wall draw isn't that low after re-examining some reviews.

The numbers I gave (50W, 75W, 100W) were not scientific, they were illustrative only. My question is mainly how much power is the CPU using at say 50% load; is it 50% of the difference between idle and full load i.e. linear? Or is it some quadratic function?
 
i would say its generaly linear, though theres no definitve answer because different loads use more/less energy. i can load my cpu at 100% with folding@home and not use as much power(or create as much heat) as loading it with orthos or TAT would do. even TAT puts more stress then orthos.

you can loosely judge power consumption by looking at your temps. if one program pegs your CPU at a higher temp you can usually assume its using more watts. this method is obviously not very precise.
 
That's a good point, never considered that different loads draw more or less power. TAT generates very high temps, but I have no clue what it's doing and I don't think anyone else does either, whereas Orthos is calculating stuff... weird.

Can someone with an engineering background or electrical background chime in on this? Though, you're probably right, since P = IV then the draw could very well be linear (although there is that voltage droop at full load, so, who knows).
 
spotpuff said:
20W swing from full to idle seems low, but I was more interested in moderate usage, say 50% CPU usage, how does that power draw compare?

20W is low compared to what I've seen on Tech Report's reviews of CPUs at idle vs load; typically the swing is 50W+ and sometimes much higher.
Click to expand...

It seemed a bit low to me too, but those are the numbers I see, on an E6400, oc to 2.66GHz but at stock (1.3v) volts, running Orthos as the loader (TAT would likely cause it to draw more).
 
ghost6303 said:
you have to remember what watts are what.
watts= power draw (amps) x current (volts)
if your video card uses 100 watts and runs at 2 volts, then its using 50 amps.

but now your household power is 120v. (or 240 for you damn foreigners, but say 120 for simplicity) if you use 20 watts at 120 volts, thats a little less then 0.2 amps. that 222watts your reading is your household current, 222watts is around 2amps at 120v. but at 12 volts (after the power supply) that would be around 19 amps.

if any of that made sense, it illustrates why a processor rated at 100watts from the power supply does not mean 100 watts from the plug in the wall.
Click to expand...


I don't really get what you're trying to say.
Watts are watts, energy per time unit. 100W power draw by the CPU equates to more from the wall socket only because of inherent losses from inefficient voltage conversion.
Watt = unit of power
Amp = unit of electrical current
Volt = unit of electric potential
 
spotpuff said:
That's a good point, never considered that different loads draw more or less power. TAT generates very high temps, but I have no clue what it's doing and I don't think anyone else does either, whereas Orthos is calculating stuff... weird.

Can someone with an engineering background or electrical background chime in on this? Though, you're probably right, since P = IV then the draw could very well be linear (although there is that voltage droop at full load, so, who knows).
Click to expand...

Unfortunately, semiconductors are not necessarily ohmic, so they don't follow Ohm's law.
 
Argh, I had a HUGE reply typed up, but firefox crashed.

That's it, I'm going back to IE.

if any of that made sense, it illustrates why a processor rated at 100watts from the power supply does not mean 100 watts from the plug in the wall.
Click to expand...

100w is 100w. If the CPU is drawing an extra 100w from the 12v line, it will be drawing an extra 8.5a from the PSU. That 8.5a from the PSU will require at least .85a from the wall... A bit more actually (as most PSU's lose 25% efficiency during normal operation).

If a CPU requires 100w more than another CPU, it WILL take 100w extra from the wall too.


Also, it's important to note that as CPU load increases, NB and RAM loads generally increase too, as do the FSB and Vreg loads, all which contribute to the majority of the "load" power you witness.

The difference in wattage between HLT and INIT in older CPU's (with higher TDP values) was closer to 40w (stock speeds of course), so, 20w doesn't seem all THAT far fetched.
 
Arcygenical said:
Argh, I had a HUGE reply typed up, but firefox crashed.

That's it, I'm going back to IE.
Click to expand...

Meh, in Vista, IE crashes a heck of a lot more for me!

And yeah, I thought everyone knew that power-draw from the wall would be higher than what the power supply is able to give out... don't they explain that in every [H] power supply review? meh...

CPU's are pretty easy on watts though. Now it's just the video cards and 10 hard drives you have to watch out for.

So in direct reply to the OP... I dunno...
 
You must log in or register to reply here.
Back
Top