Testing Your PSU Rails

[TheCleaner]

Testing Your PSU Voltage Rails
Introduction

It is crucial that you know your TRUE voltages on the 3 main voltage rails, present on all ATX PSU's.
Many people still rely on means of software, such as MBM, Speedfan, and the BIOS, but these are all very unreliable. I have been able to prove this, and what it brought to my attention, that those who trust software, are potentially putting their software in danger, particularly those who have adjustable rails in their PSU.

Why NOT to trust Software

I own an Antec Truepower 550W PSU, which is easily capable to handle any desktop system around. It would not struggle at all to power my system

AMD Athlon XP Mobile 2500 overclocked to 200 x 11, 1.6vcore
2x 256MB Corsair XMS4000, 2.6vdimm
ATi Radeon 9800pro, overclocked to 412/365
Single 120gb WD 7200rmp HD
Sony CD-RW/DVD-ROM, 52x32x52x/16x
Volcano 12 CPU cooler ran off standard molex @ 4200 rmp
1x120mm fan, 2x80mm LED fans, running from 'fan-only' connectors

Though according to various 'in windows' monitoring applications and even the bios, this was not the case;

Various Software Readings:



And the BIOS:



A common assumption is that the bios will always give the correct voltage, this is not true at all. I bought myself a multimeter, and tested these rails, and these were my findings;



Surprised? Well there’s the hard proof, software/bios is not accurate.
Now think of the consequences of someone adjusting their rails up to spec according to software, they are going to seriously jeopardise their components.

What to do Now?

So what do you do about this?, well do what i did, buy a multimeter or voltmeter. They are very easy to come by. I got mine off eBay for a very good price. Check DIY stores too, but they do usually overcharge.

Though, it is important that it’s also got a certain degree of accuracy. My multimeter, when reading at 20V DC, has a resolution of 10mV, with an error of 0.5%. Since we are testing voltages above 1V, the resolution and error is perfectly acceptable. Find yourself one. Here is an eBay search for 'digital multimeter', as you can see, theres plenty, and going for under $10, its an investment well worth the price.

Preparation

OK, so you have got the multimeter at hand, time to see what your rails are REALLY like.

CAUTION: This test involves using probes with the computer ON. Please only do this if you feel competent enough to be dealing with electricity. If you short rails, you could risk damaging the PSU and possibly your components. I cannot be held responsible if something goes wrong, this is merely a guide. If you feel any concern, please post BEFORE doing this, and I will be glad to help.

Firstly, you will have to set up the multimeter to read the voltages you will be measuring. You should be able to move a dial on the multimeter. You want to move this to the number 20 (like this), in the section which relates to DC voltage. This is usually represented by THIS symbol.
Put The Black probe into the 'COM' plug on the multimeter. The Red probe should be placed in voltage detection plug.
DO NOT insert the probe into the 'current detection' plug, which on cheaper MM's usually has a '10A max or 10ADC' Label.

NEVER ATTEMPT TO MEASURE THE CURRENT/AMPERAGE OF YOUR PSU

You HIGHLY risk frying your PSU, and any components connected to it, your entire computer basically.
ONLY people with a high competence for electronics, and a very good industrial/professional multimeter/ammeter should ever want/need to

Now, while the PC is OFF

- Locate a free 4 pin Molex Connector, although you can test the back of molex connectors, if you have non spare, it just might be a bit more akward.

- Locate the AUX connector



And pull these into an area free from any clutter, so they are accessible for the probes on your multimeter. This may be a good time to do some cable rearrangement.

Measuring the Voltages

Time to start. You want to observe these 3 voltages in all states. So you will have to set up the probes BEFORE you turn the pc on, because you want to observe the voltages for.

- POST
- Boot
- Windows Startup
- System at Idle
- System at Load
- Shutting Down

That is the problem with most PSU measuring guides, they test the PSU externally, with only the load of maybe a few fans, which does NOT give any real kind of indication on how the PSU copes powering a computer. That’s why my guide shows you how to test with the PSU under different levels of load.

So, set up each of these measuring, and observe/record the voltages at each stage as listed above. Ideally, they should not go above or below the rated voltage, though no component is perfect, which is why there is an ATX specification, which sets parameters for each voltage.



It is my opinion though, that any PSU which falls close to the min ATX spec, needs replacing.
Another Purpose of these rail tests are to find out how much your rails fluctuate, and if they do a lot, even if they stay within ATX spec, id suggest replacement. Clean, Stable voltage is just as Important as enough Voltage.

Measuring The 12V Rail

For this you just need the 4-pin Molex. Insert the red probe into the yellow 12v connector, and the black probe into any black Ground connector on the Molex:



Measuring The 5v Rail

For this you just need the 4-pin Molex. Insert the red probe into the red 5v connector, and the black probe into any black Ground connector on the molex:



Measuring The 3.3v Rail

For this you will need the AUX Connector Only, or a combination of the AUX and 4-pin Molex.
Insert the red probe into either one of the two orange 3.3v connectors on the AUX connector
Insert the black probe into any of the black ground connectors on the AUX connector, or the 4-pin Molex connector.

Note: Im aware that people may struggle to insert the probe into the aux connector like is pictured, but you can also test the rail via the back of the connector.

Using Just AUX Connector:



Using AUX and 4-pin Molex:



Other means of testing:
Driverheaven have put together a very good guide too so please give it a read, its shows you with pictures how to test from the ATX connector
Driverheaven's PSU Rail Guide

What Did You Find Out?

Do you need a new PSU, is your current PSU a lot better than what software had you thinking, post what you found?

If you think you need a new PSU, consult davidhammock200's thread HERE

I hope this guide helps. If you have any suggestions, found any errors, need more info, I will edit/add to this. Feel free to use the images; they are all hosted at imageshack. Alert me if you get red X's and I'll sort them.

although it has little bearing over how well the psu performs with a load, you can test the rails of a psu before you put it insid ethe computer, by manually starting it

see this thread for instruction
http://forums.extremeoverclocking.co...d.php?t=118964

its a good way to ensure a brand new psu wont fry your components because its been damaged in transit for some reason, the rails should all still be in spec, and if they arnt in spec with no load, they dont stand a chance with load.

[davidhammock200]

Please keep going!

[Verbitan]

Brilliant! What else is there to say? I vote sticky although i am 99% positive that it will be made a sticky anyhow! Great work...

[dr_sharp]

This thing was born sticky.

[Nitrix]

nice work man...helped me w ma new comp

Voted Stickey

[Transnone13]

Good job, so true. I just double checked mine and my software is WAAAAY off. I knew my 510 Deluxe wasn't dogging it.
Software says:
11.87 actual = 12.15
4.98 actual = 5.15
3.28 actual = 3.35

All this while under dual instances of folding.

Many people say the same things, but you backed it up with testing and picts way to go .

PS:

You may want to add you can get the voltages from the other side of the molex if it is plugged into a device.

[onewecallgod]

HAHA

i love it when people try to get stickies! he's been trying for a while now

you can sum this thread up in 10 total words

[Susquehannock]

Nice. The pictures will help a great many I'm sure. And the use of an
el cheapo DMM was a nice touch. Not everyone can go out and
buy a $200 Fluke meter just to check their voltages.

This DMM for $40* has served me very well so far. I've even checked against a couple
Fluke meters and it seems spot on. It has a temperature probe which
comes in handy. And the RS-232 interface is great as you can monitor your
DMM from your computer screen.

Keep up the good work.
Will reference others to this thread from now on.

[WiCKeD]

Great job bro - I need to do this! Maybe I can avoid electrocuting myself again...

[acidbaby]

:::subscribed:::

[Nitrix]

yo i love that post helped me out tons

[Gonwee]

My software (ABIT EQ) says 11.78-11.8V but with a multimeter it come out to about 12.04. 5V says 4.94 but is 5.03 and 3.3 says 3.25 but is 3.31

[D_unruled]

what is that AUX actually used for?

[TheCleaner]

not actually sure.. its legacy i think, something for old P4's, and i rekon they remove them from new PSU's these days.. which may present a problem, because that involves using the atx connector (the only other connector which has a 3.3v line) but, if your probe is thin enough, you will be able to test it from the back of the atx plug, hense being able to test it while the pc is on.

so thats an update to come, testing 3.3v rail, even if you dont have aux connector

[Tim]

Thank you sooo much, do you guys also feel that EOCF is one of the best places to learn? I mean we have really a lot of sticky's I think alot of forums would be jealous about these......

The Cleaner GOOOOD JOB! here dancing banana for you --->

[TheCleaner]

hehe, thanks, though its what many people have been saying for ages..
its just i thought id put words into pictures and prove that a multimeter is needed, as well as show how to actually use one.. its still in progress, i hope to clean it up a bit.. so not as much reading/rambling, but more pics... shame that theres a picture limit, i wonder if a mod can 'upgrade' certain posts to allow more pics

[bjoci]

sticky worthy

[Antillano]

Great thread, Simple and helpful. Is fluctuating from 12.00 to 11.93 to bad? I guess not since it stays in the range, or is fluctuation the problem?

[Zaphodicus]

Hi peoples.
Been having some problems with both my old Athlon XP 1900 that has been Blue screening constantly with a different cause each time when I studied the dump file. And since the back up battery circuit had been faulty for more than two years I felt it was time to upgrade. so I bought a Athlon XP 2600 Barton (10) used one of my old 512 meg ddr 2700's and purchased another 512 and slotted them into a gigabyte GA-7N400 Pro2 using my old (2yrs) Haiking 450 watt power supply and my three yr old 64meg geforce2 Ti .( fresh install of win2k on zero'd drive 80gig st380011a Barracuda NTFS).. by using the gigabyte board I have freed up the PCI slots as I no longer need to insert a LAN card or an add in ide card so I can run each of my 2 hdd's and my dvd writer and my dvd reader/burner on thier own channels.. onboard sound and etc .. infact no PCI slot are used now .. so I'd imagine that must be saving some power even though I still have the same hardware built in .. however .. the system has been very unstable .. I've played with the FSB and read a lot in the forums.. the unit stabilized somewhat when I dropped the FSB to 164 and 163 ..so possibly a timing problem ?? however it was unstable at 133 .. .. figuring it would most likely be the power supply or the video card as they are the oldest electronic components I set about researching the forums (after cleaning both) for some pointers as I don't have a spare as I "usually" would and its Sunday and the best deals are at my usual supplier and they are closed today.. well I got my multi meter out after reading this article in this forum and tested the values

below is what was displaying in everst home


CPU Core 1.71 V

+2.5 V 2.59 V

+3.3 V 3.25 V

+5 V 4.27 V

+12 V 11.71 V

+5 V Standby 4.65 V

and the CPU and 12-volt rail were fluctuating quite a bit according to everest... but upon measuring with the METER I found

5 volt 5.19


and constantly fluctuating to 5.20, which should be OK

12 11.78 fluctuating to 11.79 .. seems ok .. hopefully these PSU's supply enough Wattage on the 12-volt rail.. and


3.3 33.35 volts stable .. and all values were stable at any point whether booting running hard or shutting down.

also Everst showed the front side bus and CPU to be wildly fluctuating .. sometime the FSB was 550 or more giving a total display for the CPU of 91300 and sometimes over 100000 which can't be right


So the multi meter gave a rather different report to the software report that everest gave.. I'm imagining if the 12 volt rail is "coping" wattage wise (can't find any information on the net about these haiking PSU's and there’s nothing on it to indicate what it should be rated at for any rail .. though it says its a P4 supply and it does have protection built in) so now I'm going to try an old pci S3 card in it and see how it goes .. maybe I'll get a better indication whether it’s the video card or not .. though the S3 will draw much less power and relieve the PSU somewhat .. the problem really shows while I'm capturing video through the firewire .. It’s much slower than my 1900 on the asus board it was on .. I’ve messed with the ram .. trying one stick at a time and tried my old CPU in this board with no luck .. still the same ..

memtest86 shows the ram to be ok ..

It’s just awful unstable in windows..

on some occasions when IE is launched the hard drive will thrash and it can take a whole minute for any actual activity to show on the monitor .. if I’ve clicked anything else while waiting… then everything will happen at once.. ..

explorer will crash and I'm not able to restart it as task manager won't load .. there’s no problems showing in the system hardware and its showing just over 50degres while hammering it capturing video and operating other programs while I’m trying to get some idea of what’s going on by stressing it .. I have a very large fairly new heat sink (copper) and fan with high quality thermal past applied thinly and it runs several degrees lower than with the fan that came boxed with the cpu.. air flow is increased by 2 large fans in the case, which are powered by a separate 12 volt supply not by the PSU .. it was showing these signs while only running bare minimal hardware .. no cd or DVD drives and only one hard drive and not pushing it overly hard at times .. and it will completely fall over if left for a period.. I've looked and cleaned and investigated .. so wish me luck while I try the pci video card .. though this may not nail the problem down I'm going to try.. it is stable if you let it alone while windows loads and don't push it to hard .. but I like to push ...and even with the battery circuit dead my previous machine was stable before it started blue screening…. and I pushed it … I’ve disabled everything possible from starts etc and run it offline without firewall and virus scanner .. original bios which I've read is better than the revision.. drivers that came with the mobo and latest detonator drivers for the video card.. no ad in cards at all hewlet Packard scanjet 5200c running on the parallel port as its crap on usb and HP DeskJet 930c windows 2k service pack 3 as service pack 4 causes my 512 meg USB stick to not work .. however with no drivers for these and with them unatached the problems remain....

any body wish to comment ??
Thank

"If I meet you .. I'll control+alt+delete you"
Wierd Al Yankovitch from the song
It's all about the Pentiums

[TheCleaner]

Zaphodicus, your rails are well within spec, and aint fluctuating much, so i cant see it being a PSU problem.. but, what are the rails like between idle and heavy load.