Electricity consumption meter - weird readings

[vader]

Ive picked myself up an electricity consumption meter.
A MS6115, same as shown here http://www.dansdata.com/quickshot041.htm

For the most part it seems ok.
2200w for the kettle, 35w for a 19" LCD, 28w for a 17" LCD, 7w for standby in both.

Those figures are pretty much what I would expect however when I put my PC on it I only get 83w (raising to 89w in TF2)
(specs - 3800+ X2, 6HDDs, 1DVD Burner, 7600GT)
Now this does usage just not seem right to me, I would suspect that it should be a fair bit higher.

This reading sound dodgy?

[TheLobb]

nope thats how much pc's use. so yes your 1200watt psu is a waste. but for some reason we need the big psu's to make them work

[dakiller]

Looks lower that what I'd guess, but PC's use far less than you'd think

[breech]

fyi, the recent thread on these meters & results is here:

http://forums.overclockers.com.au/sh...ht=power+meter

[SnooP-WiggleS]

sounds way too low, i'd figure if its a low power model x2 3800+ you'd do perhaps 120-150w DC when gaming (6 hdds even idling adds another 30w at least). add 20-25% psu efficiency loss and you should be reading at least 150-180w. Very rarely some psu's have a badly designed active pfc circuit that causes all sorts of weird harmonics, which makes it impossible to get accurate readings. whats your psu for interest?

see:

http://www.xbitlabs.com/articles/oth..._15.html#sect0

[Symon]

Computers are a non-linear load, so any el-cheapo power meter will read incorrectly. If you had actually bothered to read Dan's entire article (the one you linked to) he explains it for you.

[SnooP-WiggleS]

Originally Posted by Symon

Computers are a non-linear load, so any el-cheapo power meter will read incorrectly. If you had actually bothered to read Dan's entire article (the one you linked to) he explains it for you.

at least for the kill-a-watt power meter, it will read accurate as long as the harmonics are not too high frequency.

"
Reality:
Combination of phase shift and non-sinusoidal current.
AC power measurement in real life is not a trivial matter. The Kill-A-Watt accomplishes this task using an ASIC DSP.

The sum of instantaneous volt x amp divided by the number of samples during the integration period. The number of samples needs to be high enough to account for every part of the current waveform. So, if the display is to refresh twice a second, the chip would have to take the reading of both the voltage x amperage 500 times every 1/2 second, then divide the sum of the products over 500 to give a true average power(over a half second period). This isn't a hard process for custom made DSP.

PF = the wattage you get above/V*A average
..."

http://www.jonnyguru.com/forums/show...?t=2284&page=2

and thats exactly why the break with some psu's that generate high frequency harmonics - kill-a-watt and similar devices don't have the frequency resolution to do this. Basic capactive (passive pfc/non pfc) loads should be handled fine. I'm assuming the local MS6115 power meter operates in a similar fashion to the US $20 kill-a-watt.

[NewKleer]

while they might not be accurate, theyre not a huge distance off. people dont realise that computers dont use that much power, due to low quality power supplies, or ones which spread their capacity in a ratio (over the different rails) thats far different from average pc usage, everyone has been forced/scared into buying psu's that have 2-5 times the needed capacity.

[Seeker]

Originally Posted by NewKleer

while they might not be accurate, theyre not a huge distance off. people dont realise that computers dont use that much power, due to low quality power supplies, or ones which spread their capacity in a ratio (over the different rails) thats far different from average pc usage, everyone has been forced/scared into buying psu's that have 2-5 times the needed capacity.

Say your system uses 400w on full load. You have 2 choices, a 450w psu or a 1000w.

I'd take the 1000 anyday. the 450w is going to be close to its peak and will most likely squeal, run very hot, you won't have any headroom for future components. If you bought the 1000w it would run cooler (as it can handle 1000w of power), the rails would be stable, you'd have plenty of headroom and it wouldn't be stressing out too much.

It always makes sense to buy the biggest PSU your budget allows, from a trusted brand.

[LethalCorpse]

If you take an average reading of the instantaneous voltage*current (please don't say amperage, it gives me a headache), you will end up with zero power (same as above for wattage). I'm fairly certain this holds no matter what the PF, but it's a moot point - the numbers will be wrong. You need to take the root of the mean of all squared samples (RMS) to get a useful reading of PF or VA.
Are you sure it uses an ASIC, or even a DSP? I find it fairly unlikely. I pulled mine apart before finding (as I'd have known if I remembered Dan had already done the same) that there's not much in the way of identifying markings, and the IC is epoxied. So, I can't say for certain, but if I were designing such a device, there's pretty much no way I'd use an ASIC. An ASIC is extremely expensive to produce. For a start, you need in-house silicon design facilities, which is less than trivial. Then you need to pay someone to manufacture it for you, and the fixed and unit costs for a fabrication run are astronomical. Basically you don't do it unless your volumes (or margins) are large enough to absorb the costs of going ASIC (read: military), or if there is nothing out there in an OEM part which can do the job for you. In this case, all it's got to do is sample two waveforms at a suitable sampling rate - 1kHz, as you suggested, would be about right - and do some basic arithmetic on it, as well as driving an LCD display with the data. Any cheap, nasty 8-bit, low speed micro with a built-in ADC can do that. The only time you need a DSP is when you're dealing with high sampling rates, and you need to process high volumes of data at a high rate or with complex formulae. 2k sampling over two channels with a single multiplication and single accumulation per sample, two sqrts and a single multiplication per 2k samples is not a heavy load. Sure you could do it with a DSP, and you could do it with an ASIC, and it's possible they've done either of these. But if they did, I reckon you'd have to do one more operation - a BCD shift left on the price.

[LethalCorpse]

Originally Posted by Seeker

Say your system uses 400w on full load. You have 2 choices, a 450w psu or a 1000w.

I'd take the 1000 anyday. the 450w is going to be close to its peak and will most likely squeal, run very hot, you won't have any headroom for future components. If you bought the 1000w it would run cooler (as it can handle 1000w of power), the rails would be stable, you'd have plenty of headroom and it wouldn't be stressing out too much.

It always makes sense to buy the biggest PSU your budget allows, from a trusted brand.

A PSU running well below its rating will also be less efficient

[NewKleer]

Originally Posted by Seeker

Say your system uses 400w on full load. You have 2 choices, a 450w psu or a 1000w.

well, on the big assumption that your system actually uses 400W at full load (id say ud be lucky to hit 300W), last time i checked they made PSU's between 450W and 1000W...

[SnooP-WiggleS]

Originally Posted by LethalCorpse

Are you sure it uses an ASIC, or even a DSP? I find it fairly unlikely....

I was referring to the US$20 kill-a-watt, not the local thing i'm not so sure about since I don't own one. just going off what other people said.

For interest does the local one measure power factor? If it doesn't it may only be suitable for pure resistive loads anyways like kettles. Again to make a usable wattmeter (since lots of appliances - tvs etc now are switchmode based so are not resistive load) it has to do some sought of sample and integration method, of which its probably cheapest to do with a dsp when mass produced.

[NewKleer]

it does have a power factor rating, which is usually 100, though i did use it on my treadmill and the power factor was like 40 or 60 or something

[vader]

Just done another PC.
3200+ X2, onboard vid and sound, 2 HDD + burner

This one showed much more range in the readings.
70w when booting, 78w idle to 88w when doing various stuff in windows, jumping up to a max of 102w when playing WC3.

Back to the other PC, same figures as before with no real movement, unlike the second PC.
One think I did note however, is that this is showing a PFC moving between 92 and 96. Everything else is on 100.

I spose the PSU could be affecting the readings?
(I am a tad sceptical on the 6w difference for the 7600GT between desktop and idle)
My 7900GT is due back next week, be interesting to see how that change the power figures.