PC shutdown issue, PSU -12v at -9.5V ?

Associate
Joined
21 Jan 2008
Posts
484
Location
London
Hi there

I recently posted in the general section about PC issues I've had:
http://forums.overclockers.co.uk/showthread.php?t=18062029
My PC Spec
Corsair HX 620W PSU ( bought from competitor :( )
8800gtx ( bought from OCUK in ages ago )
Phenom II x3 720 @ 3.4 ( bought from OCUK in Apr 2009 )
Gigabyte GA-MA790XT-UD4p MB ( bought from OCUK in Apr 2009 )
4GB Corsair XMS3 DDR3 TwinX memory ( bought from OCUK in Apr 2009 )
Sunbeam Core Contact cooler ( bought from OCUK in Apr 2009 )
1 x WD 250GB and 1 x WD 750GB Sata drives
DVDRW drive

Since then I've had my PC shutting down on me unexpectedly a few times already, twice in the past hour.
What's strange is viewing the sensors via SIW.exe, the -12v is reading between about -9.4 and -9.9V , is this indicative of the PSU being the issue.
I've searched the forum and a few tech sites but can't get too much info regarding what the -12v reading means.
Any ideas ?
 
Associate
Joined
27 Jan 2009
Posts
400
Location
Chippenham
Try running a minimal system with no overclock so 1 memory stick, 1 hdd no dvd and no overclock on the 8800
If it runs stable then, it could be the psu or 1 of the items you have removed
So then you are into memtest for each memory stick in each slot for staters
 
Associate
OP
Joined
21 Jan 2008
Posts
484
Location
London
Thanks for the comments and suggestion guys.
Sorry, I did omit quite a few things. I have already run thorough memtest on both my mem sticks, also been running without an overclock since the issue started, never overclocked my 8800gtx, it runs hot enough just on standard settings.
But I think it's a good idea to run it with just 1 mem stick, just 1hdd, no dvd, will give this a go over the weekend.

Haven't got a multimeter but starting to think it may be a good time to get one now. I saw a few websites actually showing how to test your various PSU readings using this, guess it's worth a go
 
Associate
Joined
7 May 2009
Posts
343
Haven't got a multimeter but starting to think it may be a good time to get one now.
A meter built on the motherboard is what software reads. It is only intended as a monitor. It cannot measure voltages until you calibrate it with a multimeter.

-12 volts has few functions on the motherboard. To provide -12 volt supply with a load, power supply cooling fan is connected. Fan should still operate. But as the fan ages, a -9 volts would cause intermittent 'not spinning' fan. Then it spins on the next powerup leaving you confused.

What are all the numbers. If you think you know good and bad numbers from specs; you don't. Best test is measuring volts on any one purple, gray, green, red, orange, and yellow wires when multitasking to all peripherals.

Some who claim Prime95 is a good test are fooling themselves and you. Proper voltages must be measured when playing complex graphics (ie a movie), while downloading from the internet, while powering a USB device, while playing sound loudly, while searching the hard drive, while reading a CD-Rom, etc. Only then are you ready to read voltages with the multimeter. Then post those numbers here to learn about things in those numbers you did not even know exist.
 
Associate
OP
Joined
21 Jan 2008
Posts
484
Location
London
Well, just when I think I'm making some progress in tracking down what's causing this I get thrown a curveball googly doosra.

As Suggested by Johnerz I took my system back to basics, 1hdd, 1stick ram, no dvd and it ran fine for 2 days doing pretty much the same things that had caused at at least 2 a day shutdowns. So I put the other stick memory back in and within about 5 minutes the PC shuts down.
Eureka, must be dodgy memory !
So I take out that stick of memory, put back the hdd and dvd and it runs fine for another 2 days. So today I decided to start putting some serious strain on the PC to see if all okay, running Prime95 and Furmark to stress CPu, mem and GPU. Prime95 started fine and the minute furmark started, shutdown. I power the system back up and run prime 95 and then furmark and this time it runs fine, gpu pushes 90 degrees after bout 10 minutes, left it another 20 and no issues. Then go to play a game of PES2009 and it shuts down 5 minutes into it.

So first I thought it was one bad stick of memory, then GPU then back to PSU.
I did log a support ticket with Corsair last week and they have come back and said that I should RMA my PSU. It's a starting point, but my concern is that I send it in, they find nothing wrong and send it back and I'm out £15+ shipping. Secondly, it'll leave me without a working PC till I get this back, unless I buy another PSU ?
Any thoughts ? I'm starting to lean towards buying new PSU, I'm a bit accident prone and have a feeling buying a multimeter and attempting PSU readings whilst the system is live and at load is going to leave either my system or me with some permanent damage
 
Associate
Joined
7 May 2009
Posts
343
So first I thought it was one bad stick of memory, then GPU then back to PSU.
... I'm starting to lean towards buying new PSU, I'm a bit accident prone and have a feeling buying a multimeter and attempting PSU readings whilst the system is live
Ignore the fears. Nothing with the meter can cause damage. In fact, the risk from damage by removing memory was significantly higher for a few reasons.

You have two choices. Keep replacing parts on speculation as you are considering. Or get a useful reply in the next post.

As warned, Prime95 reported nothing useful. You have just as many suspects as when you started. After using a multimeter, then other tests can even report if memory is defective - definitively. What was wrong with your test. It was subjective, inaccurate - not definitive. Without numbers from the multimeter, then no other useful suggestion can be posted.

If probing with a multimeter puts anything at risk, well, even your hand is a greater threat to hardware if you remove anything. Not an exaggeration. (did you disconnect AC power cord from its receptacle before removing anything?) Get the meter. Stop entertaining fear. Get hard facts - numbers. Then have a useful reply in the next post.
 
Associate
OP
Joined
21 Jan 2008
Posts
484
Location
London
If probing with a multimeter puts anything at risk, well, even your hand is a greater threat to hardware if you remove anything. Not an exaggeration. (did you disconnect AC power cord from its receptacle before removing anything?) Get the meter. Stop entertaining fear. Get hard facts - numbers. Then have a useful reply in the next post.

Lol, yes I disconnect the AC power cord before doing any work in there, also ensure I'm not statically charged.

I'll have a look at high street store tomorrow for a multimeter. Best site I could find for showing how to test the psu with a multimeter is this:
http://pcsupport.about.com/od/toolsofthetrade/ht/power-supply-test-multimeter.htm

Is that okay or do you have a better guide ?
Also, will any multimeter do ?
 
Last edited:
Associate
Joined
7 May 2009
Posts
343
Lol, yes I disconnect the AC power cord before doing any work in there, also ensure I'm not statically charged...
Best site I could find for showing how to test the psu with a multimeter is this:
http://pcsupport.about.com/od/toolsofthetrade/ht/power-supply-test-multimeter.htm
You need any 3.5 digit multimeter. You will only use one setting - 20 VDC. A £12 meter works fine. Better meters cost more.

Ignore those PSU instructions. May cause erroneous conclusions. Too long. Too much work. Use it pictures and charts only as a reference to better understand what is posted below.

First get a significant load on the PSU - as large as you can without causing a crash. For example, play complex video graphics (ie a movie) while searching the hard drive, while downloading from the Internet, while playing a CD-Rom, while powering a USB device, while playing sound loudly, while ... Now the supply 'system' (not just a power supply) has sufficient loading for useful numbers.

Measure (and report with three significant digits) voltage on the purple wire where that wire connect to motherboard (push red probe inside the nylon connector - black probe connected to chassis).

Also measure voltage on the green and gray wires both before and when power switch is pressed. Report those numbers and behavior as switch is pressed. Do not get confused using that jumper. Do not disconnect the supply from motherboard. Do not disconnect or remove anything.

Measure voltages on any one of orange, red, and yellow wires as the power switch is pressed. Even if a voltage starts to rise, then falls to zero - report it. Even if nothing happens when the switch is pressed, still measure and report each voltage.

Measuring those voltages should take less than a minute. Report those three digit numbers here for analysis and to learn what you did.
 
Soldato
Joined
16 Jun 2009
Posts
7,664
Location
Cambridge
Why buy a new PSU, when you can have your current one checked and if neccessary replaced for free at the cost of £15 PnP? You won't find a new PSU for £15, and if you do, don't touch it :)

I gather your CPU is overclocked, and that at one point you were seeing differing behavior between one stick of RAM and another. It may be that one of your sticks is bad, and you also have another issue, and I suspect it may be a lack of voltage for the RAM or Northbridge. What speed is the RAM running at after the overclock? Is it still in spec? If not have you tried underclocking the RAM before upping the CPU speed? It's been a long time since I tried to overclock an AMD system, but I should imagine that pushing up the hypertransport might require more northbridge power (Assuming that's how the OC was done, and not a clock multiplyer change).
 
Last edited:
Associate
OP
Joined
21 Jan 2008
Posts
484
Location
London
whats the temps of your cpu?

Why buy a new PSU, when you can have your current one checked and if neccessary replaced for free at the cost of £15 PnP? You won't find a new PSU for £15, and if you do, don't touch it :)

I gather your CPU is overclocked, and that at one point you were seeing differing behavior between one stick of RAM and another. It may be that one of your sticks is bad, and you also have another issue, and I suspect it may be a lack of voltage for the RAM or Northbridge. What speed is the RAM running at after the overclock? Is it still in spec? If not have you tried underclocking the RAM before upping the CPU speed? It's been a long time since I tried to overclock an AMD system, but I should imagine that pushing up the hypertransport might require more northbridge power (Assuming that's how the OC was done, and not a clock multiplyer change).

CPU temp under load is about 29° C, at low usage it's around 22
Since issue started I've taken off overclock so it's running 2.8 standard, overclock was using only multiplyer change.
Hazardo - not that simple, as I would have to take the psu out and send it away, and from past experience RMA'ing stuff I'd be lucky to get it back in 3 weeks, so I'd have to be PCless for that time.

Westom - thanks for your input, I've got my newly bought trusty £14 multimeter and about to find out what voltages my PSU is pushing out ( or some other weird numbers ). Okay, actually I'm gonna have dinner, watch a bit of telly and then do it, If I'm not back by 10 please call Emergency services for me !
 
Associate
OP
Joined
21 Jan 2008
Posts
484
Location
London
First get a significant load on the PSU - as large as you can without causing a crash. For example, play complex video graphics (ie a movie) while searching the hard drive, while downloading from the Internet, while playing a CD-Rom, while powering a USB device, while playing sound loudly, while ... Now the supply 'system' (not just a power supply) has sufficient loading for useful numbers.

Measure (and report with three significant digits) voltage on the purple wire where that wire connect to motherboard (push red probe inside the nylon connector - black probe connected to chassis).

Also measure voltage on the green and gray wires both before and when power switch is pressed. Report those numbers and behavior as switch is pressed. Do not get confused using that jumper. Do not disconnect the supply from motherboard. Do not disconnect or remove anything.

Measure voltages on any one of orange, red, and yellow wires as the power switch is pressed. Even if a voltage starts to rise, then falls to zero - report it. Even if nothing happens when the switch is pressed, still measure and report each voltage.

Measuring those voltages should take less than a minute. Report those three digit numbers here for analysis and to learn what you did.


Right, here we go
Purple on full load was 5.04 and remained constant. Ran video in background, played pes2009, plugged in all my usb devices ( 1 external hdd, 1 x 16gb memory stick, 1 webcam, usb wireless ), ripped dvd to hdd, not sure how much else I could do
GRay - 0 before, then went to 1.56 and then 3.87 within a second and remained there
Green - Weirdest thing, couldn't get a reading off this despite trying about 10 times and pressing in the test lead quite hard and from different angles
Orange - constant 3.39
Red - constant 5.04
Yellow - constant 12.08

From the pin outlay I can see that the 3.3,5 and 12v seems fine, but what's the green meant to be and is the gray within the right range ?
 
Associate
Joined
20 Mar 2007
Posts
1,051
Dutch, here's a list of the colours and what they are:
http://pinouts.ru/Power/atxpower_pinout.shtml
Gray (that's grey in the uk, not to be confused with gray coding, which is correctly spelt gray) is the power ok signal, nothing serious with it taking some time to become active, but, it is supposed to reach 5V. It might be borderline with power on conditions of the motherboard and is dipping, causing your sudden switch off.
Green, couldn't get a reading? Not sure what you mean, something must be read? I'll assume you had zero, which is the usual.
How did you go about getting full load?
Do you have access to an oscilloscope? There may be more going on than a multimeter can show you.
 
Associate
Joined
7 May 2009
Posts
343
Right, here we go
Purple on full load was 5.04 ... GRay - 0 before, then went to 1.56 and then 3.87 ... Green - Weirdest thing, couldn't get a reading off this despite trying about 10 times ...
Orange - constant 3.39
Red - constant 5.04
Yellow - constant 12.08
Green would be above 2.0 volts before power switch is pressed. Drops to near zero (less than 0.8 volts) because the power supply controller saw the power switch pressed. First look for the above 2.0 volts before power switch is pressed - (to confirm good probe contact). Then press the power switch to see a voltage well below 0.8 volts.

Purple wire voltage is why the AC power cord must be disconnected before changing anything. And is power for the power supply controller. That voltage is good - well above 4.87 volts. If power supply controller is causing power off, it is not due to unstable or low purple wire voltage.

Gray wire must rise well above 2.4 volts. It reports to the power supply controller that power supply sees stable operation. Once this Power Good signal exists, then the power supply controller permits CPU execution.

Red, orange, and yellow wire voltages are ideal (if those are properly fully loaded). In fact, those voltages are slightly high suggesting you did not have much loading or that the power supply is much larger than needed. Normally, at least one voltage is slightly lower because that voltage is heavily loaded - the others less so.

Bottom line, the major voltages are stable (assuming full loading) - well within good operational parameters. Only question remaining - does green wire drop to well below 0.8 volts? A voltage at or too close to 0.8 volts means noise might intermittently shut down the supply.

If the green wire does drop to well below 0.8 volts, then the entire power supply system is OK - time to move on to other suspects.

Very few items can actually cause a complete power shutoff. Memory is not one of them. By power shutoff, you mean the green wire voltage rises to well above 2.0 volts. Or voltage on the red, orange, and yellow wires all drop to zero. A complete power off which is different from a computer crash.

Useful is to confirm any of these numbers when the system 'powers off' to see if it does 'power off' or just 'crash'. Crash means power remains on but computer stops working.

Very few items can actually cause a computer crash. A short list is memory, sound card, video controller, unstable power system (no longer on that list if those measurements were taken with full loading), CPU, and some motherboard functions.

A few questions above on what or how those power 'system' measurements were taken. Then move on to define either a 'power off' or a 'computer crash'. IOW recreate the failure to learn what the numbers report. Does power stay on?

Some computers from the more responsible manufacturers also provide comprehensive hardware diagnostics. As we move on towards the suspect, that diagnostic would be helpful. If not, we must obtain diagnostics from the component manufacturer or third parties. Search for diagnostics for any suspects on that short list. Once the power system is known stable, then diagnostics become an informative tool.
 
Last edited:
Associate
OP
Joined
21 Jan 2008
Posts
484
Location
London
Thanks guys, I appreciate your time !

Let me just clarify my issue - my pc randomly shuts down. It doesn't BSOD or hang or crash, it's literally as if somebody has switched off the unit at the wall, no fans remain spinning and only way to get PC back is to turn it back on again.

Westom, I think I took your post very literally and as such only measured orange,red and yellow at startup and once windows had started I took it off, so it would've had 0 load.

Measure voltages on any one of orange, red, and yellow wires as the power switch is pressed. Even if a voltage starts to rise, then falls to zero - report it. Even if nothing happens when the switch is pressed, still measure and report each voltage.

Measuring those voltages should take less than a minute. Report those three digit numbers here for analysis and to learn what you did.

Less than a minute, took me nearly 30minutes, as my PC is between the wall and desk leg, I have to disconnect it and move it away to connect, and I constantly had to reread your post to ensure I get the right things measured.

To get full load I did this:
Ran video in background, played pes2009, plugged in all my usb devices ( 1 external hdd, 1 x 16gb memory stick, 1 webcam, usb wireless ), ripped dvd to hdd, not sure how much else I could do
I'll do the same and remeasure the orange,red,yellow and try to get something on that green cable, good to know what it should do as I thought it should be 0 and then rise, but much easier if I know it has to be around 2 and then drop when I power on

So, I guess it's back to the drawing board ( or measuring board ).
 
Associate
Joined
7 May 2009
Posts
343
So, I guess it's back to the drawing board ( or measuring board ).
Notice you are also learning what the computer does as you trace its problem.

OK. Assuming those numbers are correct, we are seeking something that tells the power supply controller to power off the system. Most of those functions are located around the large peripheral chips that talk to memory, the CPU, and IO bus. Things that can request that shutdown include a leaking or intermittently shorting switch wire, temperature monitor for the CPU, some unique BIOS functions, and some plane old defective electronics. Visual inspection in those areas for electrolytics with bulging tops may also be a useful symptom.

Heat is diagnostic tool. Defective semiconductors that work at room temperature can also fail at the perfectly normal 100 degree F temperature. Another tool useful to isolate defects is a hair dryer on highest heat settings. Every component inside a computer must work perfectly fine when heated to temperatures that are uncomfortable to touch but do not burn skin. Another tool to find the defect is to selectively heat suspect items to recreate the power off.

Not on the list of devices that can cause a power off include disk drive, CD-rom, sound card, video controller, keyboard, memory, and NIC. Concentrate on fewer items that can cause this failure. IOW the problem is limited to but a few suspects.

Useful at this point - especially used in conjunction with heat - would be the manufacturer's comprehensive hardware diagnostics. Diagnostics are for problems such as yours. HP and Dell are the few who will provide what everyone else has - but will not share.

Above assumes the power supply system first is 100% good according to those measurements.
 
Associate
OP
Joined
21 Jan 2008
Posts
484
Location
London
Okay, redid measurements and below is the results

Green - Was 4.92 prior to power turning on, after power on it drops to 0.02, so seems fine according to previous statements

Under load which I tried as before, but also did seperate readings under a load by running prime95 Large ffts and a 1920x1200 16xMSAA Furmark just to make sure
Orange - Startup, 3.39, under load, still 3.39
Red - startup 5.04. under load still 5.04
Yellow, startup 12.08, under load went got readings of 11.98 and 12.04.

I guess from the above that it seems it's not the PSU then.
Whilst doing this testing I did a visual inspection and couldn't find any bulging tops.

So from your items to tick off it kinda leves the CPU, GPU and MB ? All but one of the shutdowns I've had has happened whilst either gaming or running a gpu stress test, reckon I should start hairdrying it a bit and seeing if it goes, although Furmark pushed it to around 90 degrees and it was still fine.
For both GPU and CPU I can surely run burn in tools like prime and furmark for longer to see if this cause shutdown, but I feel this would still not definitely point to the culprit as the MB always comes into play.
You mentioned diagnostics, as mine is custom built I obviously don't have any, but we use Hp machines at work, could I get diagnostics for that that would work/test my system ?

What is strange is that during both today and yesterdays tests the PC didn't once shutdown, and that's makes this issues so frustrating is that I can't readily at will replicate the shutdown.
 
Soldato
Joined
19 Apr 2003
Posts
13,513
Interesting thread...

What is strange is that during both today and yesterdays tests the PC didn't once shutdown, and that's makes this issues so frustrating is that I can't readily at will replicate the shutdown.

When you next have your head stuck in your case give the main power(ATX) connector a slight wiggle and see if it shuts down the computer.

It’s not uncommon for the copper connectors to splay overtime due to repeated connection or just poor manufacturing. This can manifest itself in random shut downs or 'on demand' if the case is vibrated or connector is wiggled – see link. If it does then it may be that this is your problem and if it is it’s easily rectified with a pin/needle/small screwdriver. (or you could just look at the connection and see if the connectors have become visibly splayed.)

It’s a long shot but not beyond the realms of possibility – and the fact that you’ve been pressing on the very connector (which in turn may have established better contact) when testing the PSU, and it’s not crashed, 'may' be significant.

Good luck with it.
 
Back
Top Bottom