Measuring power usage of computer

[cjm]

Hi .. Im currently putting together a small PC with only a 110 W power supply and basically I need to keep an eye out for how much power the computer is drawing as I add components in case it comes near to the 110 W and blows up!

I imagine the best way of doing this is to cut the cable between the AC-DC power brick and the DC-DC converter that is inside the case, and measure voltage and current.

Obviously Im not too keen on snipping the cable that arrived with the PSU so I am looking for a cable which has a male connector on one end, and a female on the other, which I can vandalise to my heart's content.

The female end can be seen in this pic, as part of the DC-DC converter:

It is located about 2/3 of the width of the pic from left to right, and 1/2 top to bottom. The plug that goes into it looks like one of those standard outputs from an AC-DC converter.

If anyone could point me in the direction of the cable I described I would be very grateful .. looking on

was fruitless as I didnt know the correct name of what it was I was looking for.

Thanks, Chris

[Mike Barnes]

Maplin A27AB, "ATX Power Extension Cable"?

[logized]

When a PSU is rated at a certain wattage, some (cheaper product) manufacturers may interpret it differently - for example, it may be the peak it can produce for a few seconds at a very low ambient temperature. Also, each DC output voltage is rated at a different ampage which could result in the PSU being overloaded by say a fast CPU or graphics card being fitted, even though few other components are installed. It may be a good idea to check-out the forums for the type of pc you are building to determine how much it can take eg.

type of socket you wanted may be is a 2.5mm socket - there are different sizes available (eg 2.1mm) so to be sure you need to measure the plug or take it to a maplin branch to check.

Dave

[cjm]

Unfortunately that will only let me check how much power the hard drive and CDRW use from the 12V and 5V rails - there are also two more cables that go to the motherboard which carry another 12V rail and 3.3V too. Therefore I need something outside the case that can measure the total power going into the whole lot. Thanks anyway!

[cjm]

The type of socket you wanted may be

That is a 2.5mm socket - there are different sizes available (eg 2.1mm) so

Thanks Dave,

Apparently, the outputs have: "Maximum current ratings:

+5v=10A +3.3V=8A +12V=10A

-12V=0.2A

+5VSB=3A"

This seems to add up to an awful lot more than 110W, so Im guessing that the

110W figure is the sustained amount, rather than peak. As it stands, the components I have chosen will work (in theory), based on my own calculations and the recommendations of others. However, I would like to upgrade the graphics card and therefore will need to know (once the whole lot is assembled) how close I am to 110 W.

This is the power supply in question:

it turns out that Im nearing the 110W mark, I will fork out for a 150W model instead, rather than trying a graphics card and having to watch it die in front of my face!

Oh, and the socket looks like the one I need, thank you.

[Tim S]

The output of the brick is 17.2V x 6.2A = 106W which sounds about right. That is then feeding a little (what looks like switch mode) converter on the little PCB. Now each of the many outputs may have an individual max amps loading, but you won't be able to use then all maxed out at once. (Not with that setup).

You got 1 hard disk? That is mostly on the +12V (+5V used for data, 12 for motor) - which is easily OK. 12V does go to the mainboard but I don't think the draw is high. Most of the MB draw will be on the 3.3V line for the CPU. Stay away from the Pentium Prescott - like the other Prescott, it's a fat hungry bugger. Slower too than the Northwood P4 in most cases.

As for the other main drawer, - the VGA card - I have no idea which lines they draw from.

I would next check the specs on the MB and the VGA card. The MB specs should tell you the max amps per line that it will need, assuming the fastest supported CPU is installed. For comparison my Shuttle SB75G2 has a 220W PSU and it's of a similar form factor to what you have - so you may need to be conservative. OTOH the shuttle is designed to take 3 drive units inc CD writer. So it will have been rated with the (and the Prescott) in mind.

What I would do is measure the current between the brick and the psu - you have only one line to break into. Then gauge the loading on the PCB by specs alone. Leave some margin. Don't overdo the VGA card (no new NVIDIA

100 Watter for you! - The one with the two supplimentary molexes to provide extra power.)

Probably suck it and see if you feel everything is OK - it will probably run unstable if you overload it so it should be obvious.

Damn nice rig BTW - did you design it yourself?

Timbo

[logized]

The PSU looks similar to one made by a company called LEX

spec on their site is very limited - it just says Max 10 Amp - also it appears that the main limiting factor is the AC brick which according to linitx.com is rated at 110w max output so the DC-DC converter would never be able to also produce that much power (you would need to know how efficient it is) A more realistic spec for a PSU is that of a Casetronic 120W unit see - you look at the manufacturers web site, they do actually quote a wattage for each voltage,then a maximum for the whole psu.

Dave

[Gordon Henderson]

I'd go with the bigger PSU anyway, but I'd also be more woried about it overheating, but I presume the back panel has a fan on it. That IDE drive is very close to the the top of the box and without forced air ventilation is going to run very hot!

And don't forget the startup current of the IDE drive - it'll suck a little bit more on startup.

Out of curiosity, whats the case & motherboard?

Gordon

[Ian Stirling]

If you can visualise the position of wires inside the 2 wire cable, then you can do this.

First, you find a suitable 12V load, maybe a 50W LV halogen bulb. Now, ideally you take two meters, and then measure the current flowing into the bulb, while measuring the voltage across the output. (using the meter and swapping it over will give slightly inaccurate figures.)

Now, take two needles, and place through the insulation (of the 12V side of the wall-wart) on one wire. Now, measure the voltage between them, and use the current to work out the resistance of the cable.

Now, given the resistance of the cable, you can use that as a shunt to measure the real current when plugged into the device.

[cjm]

A couple of months ago, I had the sudden desire to build a small PC for space, silence, and heat. My desktop machine is getting ever more powerful and is in a small room so you really notice the heat it belts out. I also noticed that I hardly ever use my PC to its full potential, and could therefore use something slower, but still keep the old one in another room for file serving etc ..

While I was looking at the small cases available, I found the Cubit 3 from

They also make a 'P4' version which is bigger, has an actively cooled PSU, but also features things I dont really need, such as SATA and support for the 3.2 GHz P4. So I chose the Cubit 3, in 'Larkspur Blue' from ... I also bought the acrylic panels to go with it.

Now I have to find components to suit!

So, the pic I used earlier was linked from Hoojum's assembly instructions - mine is still without a motherboard, which

havent sent me yet, and indeed, havent even charged me for yet. The Cubit will only take a mobo in the mini-itx form factor - 17cm x 17cm. Via sell low power, lower performance motherboards with integrated processors but I figured this would just be *too* slow for my liking, so I settled on the Insight P4 which will run an adequate Celeron 2GHz. The 'Thermal Design Power' for this is supposedly 52.8W, so there's half of my power gone already!

I can't seem to find any technical specifications for the motherboard itself, although I have emailed Insight to see if they can help me. I would imagine that it would not use any more than 20W, however, as it is based on the Via mini-itx boards and has virtually the same features.

The hard disk that I will be using is a 2.5" model rather than a 3.5", and draws around 2.4W while seeking (don't know about spinup) from the 5V rail as opposed to the 12V rail.

The CDRW that I will use is of the laptop slot-load variety, and uses around

5W at its peak.

One stick of 512 MB Crucial DDR is expected to use around 6W, and I am hoping that a couple of USB devices and a fan will come to less than 10W.

So, that gives: 52.8 + 20 + 2.4 + 5 + 6 + 10 = 96.2 W

That's cutting it mighty fine. I think I was being VERY optimistic in my earlier calculations!!

My original thoughts of being able to use a PCI graphics card as well are fading fast!! The power supply is, indeed, the LEX 110W kit. This was the biggest that the Cubit has pre-drilled mounting holes for. I think I may forget about this now and use the 150W supply instead, although I do not know who manufactures it, and will need to drill some holes myself to keep it secure:

will be interesting to see how close the power actually is to my calculations in the end!

Thanks again for the help guys!

[cjm]

Wow .. I totally forgot that the cable would have an impact on my readings! Time to dig out some old text-books I think! haha

[James Salisbury]

You could use a Hall Effect current probe, this will clamp round one of the wires and measure the current flowing... ask in the Electrical Engineering dept at Durham

[Dorothy Bradbury]

That is your peak current draw. Thus you are ok with that setup & the proposed PSU/DC-to-DC.

A Dell Inspiron1100 runs 2.5"-HD, TFT, Cel-2.2, SlimCD, PCMCIA, AND must also charge the battery at the same time out of not much more.

If you plan on using this as your main machine: o Consider a 2.5" 7200rpm drive - since 2.5" drives are a notable I/O bottleneck o Their current draw & thermal output isn't dissimilar to lower power units.

Laptop bricks come in 110, 120, 130W and 170W versions. There are (actively cooled) 190W & 230W versions coming too.

[cjm]

Oh dear. I have just received an email from Crucial (the RAM) manufacturer, saying that their 512MB module uses 21W. This has definately thrown a spanner in the works as now the power has risen to 96.2 - 6 + 21 = 111.2W. If I then add on a 10% margin for error, this gives 122.32W. Too much.

So, I would imagine that a higher rated AC adapter would not be enough to fix this - after all, I haven't got a clue how far I could push the DC-DC converter. The individual rails add up to a grand total of 212 W, but it could ever manage that much then I suspect they wouldnt have called it a

110W supply.

Im thinking now that I will plump for the 150W version from here:

can forsee difficulties when mounting this in the case, however, but that won't be anything that a little drilling/glueing/taping can't fix!

Oh, and the hard drive I have chosen is the 40Gb Seagate Momentus 5400RPM with 8Mb cache. It won't perform as well as a 7200RPM drive, but Im sure it will suffice for what I will use it for. This drive also has incredibly low acoustics, and uses as much power as a 4200RPM drive, so I think it is the best for my needs.

[cjm]

Very clever .. measuring the current without having to interfere with the cables. I may pop along one day and politely ask them if I may use one. Thanks for your suggestion!

[Owain]

"cjm" wrote | A couple of months ago, I had the sudden desire to build a small PC | for space, silence, and heat. My desktop machine is getting ever | more powerful and is in a small room so you really notice the | heat it belts out. I also noticed that I hardly ever use my PC to | its full potential, and could therefore use something slower, | but still keep the old one in another room for file serving etc ..

Use a diskless workstation on the desktop with Linux Terminal Server. Can run MS apps through WINE, and even Windows OS using WMWare etc.

Owain

[Ian Stirling]

21W in a RAM module without a fan will damn near cause it to melt. My entire laptop can hit 30C above ambient on about that power.

[nightjar]

Keep it in another room and use extension cables to run the keyboard, monitor, mouse, printers, external CD etc. That is how I moved all the computers, with their associated heat and noise out of my office.

Colin Bignell

[Dorothy Bradbury]

That 512MB module figure is not correct. o DDR2-667Mhz 1GB are a nominal 10W o Even using lower density modules, 21W is tough going

Note that 10W is bad enough: o Most motherboard memory VRMs will tend to struggle as 30W is neared o BTX was as much about memory-VRM & CPU VRM cooling

If Crucial were right then 4x 512MB (2GB) would dissipate over 80W. Might be worth emailing or faxing their USA counterparts for the figure.

The 5400rpm Seagate Momentus aren't dissimilar in performance to the 7200rpm Hitachi, and are somewhat cheaper so no worries there.

There are plug-in adapters to measure the power of an appliance, taking into account the power factor, but in this application calculation is ok. I suspect a desktop 3.5" drive will not fit in that case, but if it does then I might be tempted to uprate the PSU to a 120-150W version.

[cjm]

manufacturer,

I am glad other people have doubts about the 21W figure - Kingston DDR, although a different manufacturer, must be pretty similar, and is rated at

6W.

However, this is still pretty close to the 110W so I think I will definately be upgrading to 150W once I get the chance, in order to power a PCI graphics card too and play some remotely recent games. I would also still like to be able to measure the power that is going into the computer though, to see just how accurate all of these calculations have been!