• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

[GUIDE] Measuring computer power consumption using multimeter

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

cyberfish

Member
Joined
May 23, 2008
Location
London, England
Disclaimer: I am just a second year electrical engineering student, so there could be safety precautions I'm not aware of. Even though the procedure should be very safe, playing with mains power (or any kind of electricity) has some inherent risk, so follow at your own risk.

Want to know how much power your computer consumes?

Kill-A-Watt is nice, but it costs quite a bit, and you can't use it for anything else.

With a clamp-on multimeter, you can achieve pretty much the same thing (I will explain what's the difference later), and a multimeter is a useful tool to have around anyways.

However, unless you want to be poking around main's power with bare conductors (generally a DANGEROUS thing to do unless you have appropriate training), you need a multimeter with a clamp. Commonly referred to as clamp-on multimeter or clamp multimeter.

Something like these
http://www.multimeterwarehouse.com/clampmeter.htm
(first Google result, I am NOT recommending buying from them, and I have never heard of them)
I am not sure if anyone makes clamp test leads for regular multimeters, but that's possible.

They are like regular multimeters, except they allow you to measure AC currents without making contact with the actual wires (using electromagnetic effects).

I happen to have one lying around, so I decided to give it a try.

The clamp meter is very easy to use. Just set it to measure AC current with the correct range, and clamp it around a wire.

Problem is, a computer power cord (from outlet to power supply) actually has 2 wires (we can ignore the ground wire here, since it shouldn't carry any current), with current flowing the opposite ways. So, if you clamp the meter around the power cord, you should see a 0, since it measures the sum of the currents in the 2 wires (they are equal magnitude, opposite signs).

Therefore we need to strip it first, to allow us to clamp it onto only 1 wire.

First, unplug it from BOTH ends!

Stripping can be easily done with a pair of scissors. Remove a section of the outer insulation. Try not to damage the 3 wires inside. Carefully examine the wires afterwards. Use electrical tape to fix them if you damaged the insulation of the wires, like I have done. If you actually cut the inner wires in half... I guess you can fix it by stripping both ends, tying the ends together, and wrap everything in electrical tape. Add a drip of solder if you want. Not really important. Make sure there is no exposed metal!
IMAG0002.jpg
IMAG0003.jpg

With the power cord still unplugged, clamp the meter around ONE of the three wires. 2 of them should give you identical values, and 1 should give you 0 (ground, if it's not 0, run for cover). Usually, brown/black is live, blue/white is neutral, and green/yellow is earth ground. Either live or neutral will do (they will give you the same value, since one is the return path of the other).

Set your meter to measure AC amps (range should be ~0-10A, 10A is ~2400W), and turn on the computer.

Now the reading on the meter is how much current your computer is drawing at that instant.

The "apparent power" is voltage multiplied by current. Voltage is 120V is North America, and 100-130/200-240 elsewhere in most places. Current is what you are seeing on the meter.

For example, if you are in the US (120V), and you are seeing 1.5A on the meter, your computer is consuming approx. 180W.

However, the "real" power draw is not the same as apparent power. It's less. That's the power factor due to the voltage and current waves being out of phase. But since all modern power supplies have some kind of power factor correction (PFC), they should be pretty close. Kill-A-Watt has better accuracy because it can measure the real power.

Also, this is the input power to the power supply. Therefore, to calculate the power your computer is drawing, you need to divide it by the efficiency of the power supply. There is no easy way to measure that, so you'll have to rely on reviews. For my EarthWatts I'm assuming 80%. So if I measure 100W from the wall, that means my computer is drawing 80W.

For my computer (E6300 @ 2.8ghz, 9600 GT heavily overclocked), I get

Idle - 1.13A (135.6W, 108.5W out)
Orthos - 1.59A (190.8W, 152.6W out)
Orthos + Furmark - 2.32 (278.4W, 222.7W out)

So the 380W EarthWatts I am using is much overkill for my machine :).

Imagine what kind of machine would actually NEED 500W. I think we really should stop buying larger and larger power supplies just because people make them. Very few people actually need 500W. I would guess 650W would do for 2 GTX 280's (236W max power) in SLI, with an overclocked i7.
 

Chixofnix

Member
Joined
Nov 18, 2003
Location
Kansas City, KS USA
Thanks for the writeup!

I would be curious to ask you to be an independent 3rd party to help verify something...

The following link seems to be a fairly "realistic" power supply sizing tool for new builds/upgrades. Most "PSU calculators" out there, when I plug the components of my rig below in, suggest something in he 450W+ range, where my actual wimpy-capacity PSU does the job quite nicely (and has done so for some years now, thanksverymuch :))! I think this has to do with the majority of such tools being funded/hosted by PSU manufacturers/resellers out to make a few extra bucks :screwy::sly:...

Would you mind plugging your components into this tool and tell us how close its recommendations come to your measurements?

http://extreme.outervision.com/psucalculatorlite.jsp
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
Very interesting write up!

A couple of things I wasnt sure about..

1. I always heard that b/c of aPFC that most PSU's have that tends to render the KillaWatt more inaccurate than a DMM.
2. As far as your PSU being overkill for your system. I disagree completely. In fact I would venture to say its perfect. You do not want to run a PSU more than 80% of its capacity or so as its fan will likely kick up to be annoying among other things. I have heard a 60-75% is the sweetspot. You should always have adequete headroom for future upgrades and transient response. I dont think you can buy a mid-high end GPU and throw it in that system for example.
 
Last edited:

RollingThunder

Destroyer of Trolls & Spammers
Joined
Jan 7, 2005
Very interesting write up!

A couple of things I wasnt sure about..

1. I always heard that b/c of aPFC that most PSU's have that tends to render the KillaWatt more inaccurate than a DMM.
2. As far as your PSU being overkill for your system. I disagree completely. In fact I would venture to say its perfect. You do not want to run a PSU more than 80% of its capacity or so as its fan will likely kick up to be annoying among other things. I have heard a 60-75% is the sweetspot. You should always have adequete headroom for future upgrades and transient response. I dont think you can buy a mid-high end GPU and throw it in that system for example.

ED,

I try to buy and run a power supply on near the same percentages you stated (~60% best guess) because:

1. I don't want the fan RPMs ramped up so I can hear it.

2. And I don't want the added heat that goes with it especially with a bottom mounted power supply.

So far neither of my HX520 Corsairs have increased cooling fan RPMs and I know I can run 9800GTs in SLI easily if I wanted to.
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
520HX as Im sure you already know, was a very good and possibly under rated PSU. I have a funny feeling the card used in this testing is either a low end GPU or an integrated solution. IIRC, even the 9800gt was still a 100W+ card. Antec EW 380 only has 28A across the 12v rails. Plenty for any quad and low end gpu. HX520 has 40A and can handle any quad and any single GPU you throw at her. 520HX is arguably one of the best at the time and at its wattage level, no doubt! :)
 
OP
C

cyberfish

Member
Joined
May 23, 2008
Location
London, England
Thanks for the writeup!

I would be curious to ask you to be an independent 3rd party to help verify something...

The following link seems to be a fairly "realistic" power supply sizing tool for new builds/upgrades. Most "PSU calculators" out there, when I plug the components of my rig below in, suggest something in he 450W+ range, where my actual wimpy-capacity PSU does the job quite nicely (and has done so for some years now, thanksverymuch :))! I think this has to do with the majority of such tools being funded/hosted by PSU manufacturers/resellers out to make a few extra bucks :screwy::sly:...

Would you mind plugging your components into this tool and tell us how close its recommendations come to your measurements?

http://extreme.outervision.com/psucalculatorlite.jsp

It suggests 246W for me (90% load, 30% capacitor aging), so I'd say it's fairly accurate.

1. I always heard that b/c of aPFC that most PSU's have that tends to render the KillaWatt more inaccurate than a DMM.
I am not sure what's the theory behind that and it certainly doesn't make sense to me. Active PFC (and passive PFC, although to a lesser degree) makes the load of the power supply behave more like a linear load, meaning the "real" power and "apparent" power will be more similar. In any case, a Kill-A-Watt can measure both the apparent power and the real power, and a DMM can only measure the apparent power (actually, the current).

2. As far as your PSU being overkill for your system. I disagree completely. In fact I would venture to say its perfect. You do not want to run a PSU more than 80% of its capacity or so as its fan will likely kick up to be annoying among other things. I have heard a 60-75% is the sweetspot. You should always have adequete headroom for future upgrades and transient response. I dont think you can buy a mid-high end GPU and throw it in that system for example.
Well, fan noise is a valid point, but not for me. I have 5 computers in this room, with all fans on full-blast :) I can't really hear the power supply fan. If the power supply is rated for 380W, that means it's guaranteed to work reliably if you draw less than 380W, including 379W. If it fails at 379W that's false advertisement, so I'm sure they all make sure the power supplies can actually supply MORE than the rated power. From the reputable brands at least. Upgrades is a valid point, too, but I'm pretty sure I won't ever be rich enough to go SLI. My full-load without video card power is 190W, meaning I still have more than 190W for the video card (plus the idle consumption of my 9600GT). I'm pretty sure I will never be rich enough to get a card that draws more than 190W, too :), especially since I'm not a heavy gamer. The 9600 GT, which is more than enough for my needs (90fps on CoD4 maxed, 1280x1024), only draws about 60W.

Transient response I have not considered. But we won't be able to test that without an oscilloscope. I've always wanted to get one... they cost an arm and a leg. Most of the power draw is in the digital circuitry, though, and they don't have a spiky transient response like motors. The fans and harddrives can draw more current when they start up, but I doubt it's enough to upset the PSU. If it's rated for 380W continuous, it can probably supply >1000W peak.

I have a funny feeling the card used in this testing is either a low end GPU or an integrated solution
If you read the post, it's a 9600 GT. That may be low end for you, but high enough for me :).
 

ratbuddy

Member
Joined
Aug 24, 2007
Imagine what kind of machine would actually NEED 500W. I think we really should stop buying larger and larger power supplies just because people make them. Very few people actually need 500W. I would guess 650W would do for 2 GTX 280's (236W max power) in SLI, with an overclocked i7.

If nothing was OC'd you might get it to boot with that PSU. With an OC'd i7 and pair of 280, you're really better off going 850+.
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
I would bet good money says a Corsair 650TX (50A 12v) could handle a stock quad and two of those bad boys... Personally I wouldnt do it for long as I like runing PSU ni the 60% or so max range. I would go 750TX and be quite pleased and still have plenty of headroom.

For the record, My PC Took 463W at the wall (393.5W @ 85%) at 3.8GHz Q9450 with stuff in sig. 2 280s would smash my one gtx 260. Im with you on this one more or less RB.
 
OP
C

cyberfish

Member
Joined
May 23, 2008
Location
London, England

They are comparing Kill-A-Watt to using a multimeter to measure the DC currents INSIDE the computer. Depends on what you want to measure (power draw from the wall or power draw by your components), one will be more accurate than the other. The clamp-multimeter method is at most as accurate as Kill-A-Watt. There are inherent inaccuracies in measuring an AC current (because it's changing constantly), and that's what they are talking about. To measure DC currents, though, you will need to slice every single power wire in your computer and connect your multimeter in series...

Active PFC has nothing to do with that.
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
So Spectre and OWolf are incorrect in that thread and a KaW is accurate? Sorry, Im awfully tired tonight and for some reason that explanation didnt hit home..
 
OP
C

cyberfish

Member
Joined
May 23, 2008
Location
London, England
Accuracy is a relative thing. All instruments have error margins.

They are comparing Kill-A-Watt to measuring internal DC currents using a multimeter.

Measuring internal currents will be more accurate for the purpose of determining power consumption of various components, because PSUs have <100% efficiency. It has nothing to do with KAW vs DMM. If you use a DMM to measure the input current (like what I am doing here), you'll get the same "inaccuracy".

Active PFC is another issue entirely. From reading the thread, no one said it explicitly, but they are implying that the KAW doesn't actually measure real power. It multiplies the apparent power by an assumed power factor to obtain the "real" power. That would introduce errors with active PFC (since active PFC has better correction). In that case, the KAW is no more accurate than the DMM, but no less accurate either, since it can also display "apparent" power.