No power after sleep - bizarre problem

EuroSong · Aug 5, 2014

Hi guys (and girls),

In my time I've had more than my fair share of hardware issues - however this one leaves me completely stumped. I have a self-build computer: current hardware a few years old now, but is generally very stable and reliable. AM3 processor, 16GB DDR3, Gigabyte motherboard (can't recall the exact model while typing at work), Nvidia GeForce GTX 460.

I have Windows 8.1 Last night, for the first time on this computer, instead of shutting down I tried putting it into sleep mode (S3) as an experiment. The computer powered down and looked like it went to sleep.

However, when I tried to move the mouse or press a key on the keyboard, it wouldn't wake up. Moreover, even when I then tried the power button, there was NO response. No fans starting. No bleeps. No monitor output. No POST. Completely dead! At this point I got quite worried.

At first I thought there might be a problem with the computer's power switch. I therefore took a multimeter to the motherboard pins for the power button. It registered a trace voltage (can't remember exactly what it was now, but it was very low. Not 5V.

I then unplugged the PSU's main connector from the motherboard and shorted out pins 15 & 16. The PSU immediately started supplying power, and all rails (3v, 5v, 12v) tested OK. Therefore it is not a PSU fault.

I tried turning off the PSU at the main switch, and leaving it a while before trying again. No luck.

I tried removing the motherboard's battery and clearing the CMOS by shorting the jumper. No luck.

I tested the motherboard's battery. It seems to be fine, outputting just over 3v.

This is a really, really bizarre problem. The computer was working perfectly until I put it into sleep mode. However it now appears as if the motherboard has "set itself" to sleep mode, and does not know how to wake up - even when the power button is pressed. From experience, I know this makes no logical sense. Sleep mode is a memory state, and I see no reason why it should interfere with the actual power connection from the "on" button. Therefore it appears to be a hardware fault - but this also makes little sense, because as I said, the computer was working perfectly until I selected "sleep mode" from the Windows menu. Unless it's just an enormous coincidence that at the very second I selected sleep mode, the motherboard spontaneously failed! But that's too much of a coincidence.

Unfortunately my motherboard does not include an LED, so I can not tell immediately if there is proper power to the motherboard. However I did notice something I had not noticed before: when the PSU is turned on, then the keyboard's Num Lock LED is very dimly lit. Pressing the Num Lock button does turn the LED on and off. It is a PS/2 keyboard. This suggests that there IS a small amount of power running through the motherboard, and to the keyboard. I'm not sure if this is normal in any power state. Can someone answer this? When you put a computer into S3 sleep mode, and it may be woken from the keyboard, should there be a small current to the keyboard to allow this? Does the current also exist when the computer is off (but the PSU is powered)? I never noticed before if it did this when completely off.

I'd be very grateful if anyone could shed some light on this. Is it possible that it has anything to do with the sleep mode, or is that just a coincidence? I really don't want to have to buy a whole new motherboard if it's not necessary. Are there any particular power tests I can perform on my motherboard with my multimeter which can help diagnose the problem? What should the voltage to the PWR SW pins be?

Thanks.

InvalidError · Aug 5, 2014

You might want to measure voltage on the 5VSB power pin - everything that needs power while the computer is on standby gets its power from that pin, which is the purple wire on PSUs that follow the ATX color-coding scheme - preferably by back-probing the ATX connector to see what is actually being delivered to the motherboard.

As for the low "not 5V" voltage on the power-switch pin, 3V is still high enough to qualify as high for both CMOS and TTL logic. If it is under 3V, then it starts to become suspicious - assuming the 5VSB output checks out fine.

And yes, it is normal for the keyboard and mouse to receive power when wake-on-keyboard/mouse is enabled since your keyboard/mouse would not be able to detect and signal events otherwise.

EuroSong · Aug 5, 2014

InvalidError :

Thanks for your reply. I will do more testing tonight. However as I said, I did test the PSU's output from all pins and they all seem fine. I will specifically test the one you mentioned. But I do recall that when the PSU was "off", then there was still 5v being delivered to one pin (15 or 16, whichever was not the ground). That sounds like it would be the one you mentioned.

The power switch pins (on the MB's front panel connectors) does not even have 3v. In fact it just has a bare trace. I will re-measure it tonight to tell you exactly what it is. Do you (or anyone else) know what it *should* be? Should it be 5 (or 3)v? Or should it just be a trickle? If it should be more than a trickle, then that implies that the MB is somehow blocking the full power switch voltage from activating.

Your response regarding the small current to the keyboard makes sense in terms of the computer being in a sleep state. However therefore the mystery remains as to how the motherboard can "remember" being in this state... and moreover how it can be stuck in it, given that I have disconnected all power, removed the battery, and reset the CMOS jumper. Anyone have any more ideas?

InvalidError · Aug 5, 2014

EuroSong :

The ACPI state and memory content are maintained by the 5VSB supply. If you removed 5VSB power, the motherboard will not remember anything. If your computer refuses to power back on, you have some other issue.

The reason I am specifically suspicious of the 5VSB supply is because you never used the standby supply beyond the soft-power switch before and now, the first time you did try to use it for standby/sleep, your computer stopped working.

Another possibility would be that the motherboard malfunctioned while switching the memory from the main supply to 5VSB - you never used this function before, so you have no way of knowing if it ever worked right.

EuroSong · Aug 5, 2014

I just got home and performed further tests. While at work, I left the motherboard's battery out all day just to make doubly sure that there was nothing lingering there.

I tested the PSU's output on purple pin 9 (5VSB), and it read 4.12v. Is that within tolerance?

I then tested the voltage across the motherboard's panel connectors for the power switch. The mere trickle to which I alluded earlier is in fact 0.05v. Obviously, this is significantly less than the voltage you appear to expect.

Could you please confirm if this 0.05v across the power switch could possibly be normal for any motherboard? If not, then it certainly seems that the problem lies with the MB.

Also.. is the 4.12v from the PSU standby pin acceptable? Could it possibly be that this reduced voltage is some way results in a failure of the MB to pass on sufficient standby power, and only a trickle goes to the switch? In which case could it be a PSU problem?

By the way, even after a day of no power, the keyboard's Num Lock light is still powered when I plug it in. Yes, I understand that this is normal in sleep mode to allow the keyboard to wake the computer. But is it normal when the computer's supposedly off? I never noticed it before.

Thanks for your help so far. This is just so frustrating, not knowing exactly where the problem lies.

InvalidError · Aug 5, 2014

The tolerance on the 5VSB pin is 5% just like all other major rails so 4.12V puts you 630mV out of tolerance or nearly three times out of spec. This can certainly be a problem - the ACPI chip might be stuck in under-voltage lock-out.

For the switch terminals, the ACPI power switch is just a regular active-low CMOS/TTL input and as such, I am expecting 2.5-3.3V across them with the switch disconnected. 0.05V is definitely much too low.

EuroSong · Aug 5, 2014

InvalidError :

So what do you think? New PSU or new motherboard?
A new PSU is certainly preferable for me... But I'm worried that if I get a new one it could turn out that the problem is in the MB after all. It seems so strange that the PSU could suddenly start delivering under-spec standby voltage, after previously working fine.

EuroSong · Aug 5, 2014

Oh, I just performed another test on a whim: I used my multimeter's RESISTANCE test to check connections... and I see a connecting circuit between the computer's metal case and the power switch pins on the motherboard's panel. There is also contact somewhere between the case and the CPU heat sink. SHOULD there be any kind of grounding of the pins? Or could it simply be the case that I have a short somewhere, which is causing the. PSU not to turn on as a failsafe? And could a short be responsible for the trickle voltage?

Please let me know if there is supposed to be contact...

Thanks.

InvalidError · Aug 5, 2014

EuroSong :

The new PSU might be inevitable since 4.12V is definitely not normal no matter what might be going on with the motherboard. If you can, you could try measuring the current the motherboard pulls from the 5VSB pin to see if it is putting some abnormal load on it. Without any fancy peripherals plugged in that might draw standby power, the motherboard should be well under 1A on 5VSB. If you have more than 1A, unplug all external devices and try again. If the motherboard is still drawing over 1A, something on the motherboard might be busted and drawing excessive 5VSB power.

What PSU brand and model do you have anyway? At a glance, you do not appear to have mentioned it yet.

EuroSong · Aug 5, 2014

InvalidError :

My PSU is a NorthQ NQ-4350GP-FL

I just unplugged everything from the PSU and re-tested the standby voltage: 4.7v. I don't know if this is significant.
How can i test what current the more board pulls from the standby power?

InvalidError · Aug 5, 2014

There does not appear to be much info about your specific model but on the "PSU list," NorthQ is one of those brands with the dubious honor of getting listed as tier-5, which means the brand is known to have many problematic if not outright dangerous PSU models.

As for how to test for current, there are three ways:
1- use a clamp-style amp-meter
2- use a hall-effect current probe
3- use a pin extractor to remove the pin from the ATX connector, connect one of your multimeter's probes to the loose wire, insert the probe in the open location in the ATX connector and setup your multimeter for current measurement... if you have no pin extractor, the alternate method is cutting the wire to open the circuit and insert your multimeter there

4.7V is still out of spec by 50mV, which should be enough to make your motherboard do something if it is not dead. You might want to leave your multimeter hooked up as you try to turn your PC on to see if the voltage changes during the process. Since the main converter bootstraps off the 5VSB supply during startup, it most likely will.

EuroSong · Aug 5, 2014

Hmm. Thanks, but I don't have a clamp style tester and i really don't want to start cutting wires...

I'm thinking of trying a new PSU. I was unaware of this tiered list, but I see that NorthQ features under tiers 3 and 4 too, depending on which version of the list you look at.

I may buy a new PSU. But before I commit myself to uncertain spending, could you please give me your opinion on just how likely the problem is to be with the PSU; or how likely it is to be with the motherboard? It' be extra annoying to order a new PSU; connect it up; and discover that the problem is the same, meaning I have to buy a new motherboard after all and I just spent unnecessarily on a PSU when there was nothing wrong with the old one...!

Edit:
I just re-tested. Standby voltage on the purple wire crept up to 4.8v. Then I shorted out pins 15/16 again, and re-tested all the outputs. They're all good and within spec... EXCEPT the standby wire, which fell to 4.32v. Is it normal for this to drop voltage when the PSU thinks the computer is running?
I then took out the 15/16 short (turning the PSU off again), and the standby output dropped immediately to 4.66v. I want to know if these fluctuations are normal for a PSU.

Further edit:
Ok, this is really puzzling me. I took my multimeter again to test resistance between the grounded metal case, and each of the holes in the motherboard's connector (into which the PSU cable plugs).
For EVERY hole, there was a connection to the chassis.. Except Pin 8 ("power good"). Does this mean I have a serious short circuit somewhere? Or am I missing something? (This test was performed just on the MB and case, while the PSU was out of the equation).

InvalidError · Aug 5, 2014

EuroSong :

The whole point of tolerances is to specify a supply operating range the PSU should NEVER wander out of during normal operation. Being significantly out of tolerance is NEVER normal under any circumstances and when the voltages are as far out of specs as your 5VSB is (16-17% off when the spec requires better than 5% accuracy), you can obviously expect issues. A 5VSB supply that works normally will remain in the 4.75-5.25V range regardless of load or whether the main PSU is on/off; preferably closer to 5.00V.

And yes, you will read resistances between motherboard power pins and ground since ground pins are obviously connected to other grounded items across the PCB such as the rear IO shield and motherboard studs. The supply pins also have decoupling capacitors and various electronic devices connected to those rails that will look like shorts to a low-power source like a DMM ohm/diode-meter. If you test the PSU side, you will read 15-150 ohms resistors from the PSU's internal dummy loads.

EuroSong · Aug 5, 2014

InvalidError :

Okay, thanks. Yes I just tested the PSU's own pins against one of the grounds, and I see connections with some resistance also.

So what do you recommend I do? New PSU? New MB? Or is there anything else you would try first if you were in my position?

InvalidError · Aug 5, 2014

EuroSong :

If it was my PSU, I might try reverse-engineering the 5VSB circuitry to figure out why it is so far off and try fixing it mostly to satisfy my personal curiosity and possibly post a video or write an article about it. Considering the knowledge, skills and danger involved with working inside a live PSU, that would not be a reasonable recommendation for normal people.

Since the 5VSB on your PSU is effectively hopeless, your first step would be getting a decent quality ~500W PSU (most good quality sub-500W PSUs cost about the same as 500-650W units anyway) and those usually start around $40 when looking at tier-3 and up. The Antec VP450 is Tier-2b and costs $40-45 regular retail price.

EuroSong · Aug 5, 2014

InvalidError :

Okay... (I'm not going to dismantle my PSU or reverse-engineer anything!)
I will try a new PSU. Although I do go for the fanless models, and the one I will get is over £100 (I'm English). If I got a cheaper one and it works then I'd only have to replace it again with a proper high end fanless one.
If the new one doesn't work and I have to get a new motherboard anyway, I will be most annoyed... but this is the gamble one takes when you have no spare components to test, I guess! At least I will go for a full modular model. The one I have my eye on is a Seasonic:

http://www.scan.co.uk/products/520w-seasonic-platinum-ss-520fl2-fanless-psu-80plus-platinum-full-modular-eps-12v-atx

I hope this meets with the approval of the PSU list!

I will let you know how I get on. In the meantime, thank you so much for the time and trouble you have spent replying here to try to assist me.

InvalidError · Aug 5, 2014

Almost anything from Seasonic is a pretty safe bet.

Hopefully your wonky 5VSB was your only problem. To really test that 5VSB, the ideal thing would be to hook up an oscilloscope and look for undershoot/overshoots during the initial PSU power-on and what happens when you trigger the soft-on function. There could be a lot more stuff/junk hidden in there than what can be caught on a multimeter.

EuroSong · Aug 7, 2014

InvalidError :

WooHoo! It's working! My Seasonic PSU arrived today, and it booted right up. I did not need to replace the motherboard. I am so happy that the problem was the NorthQ PSU's standby voltage, and not a complete MB failure.

InvalidError, thank you again SO much for your help and advice. Without your guidance I would not have known what to try, and might even have bought a new MB.

The most bizarre thing remains the fact that the NorthQ PSU failed at exactly the time that I put the computer into sleep mode. How can a software instruction cause hardware to fail like this? Especially a PSU, which is just a bunch of capacitors and transformers, and does not contain any delicate chips. Totally weird.

Another mystery is that when my computer started up again, after the POST, it booted into exactly the same memory state as it was when it failed on Monday evening. I thought that the S3 sleep state (which is what I thought Windows 8's "sleep mode" was), was only a RAM state. However the fact that it booted into the same memory state 3 days after power to the RAM was cut, implies that an image of the memory was actually saved to the hard drive as per "hibernation" mode.

Anyway - these mysteries aside, the main thing is that it's working and I did not have to get a new MB and reinstall everything! Thanks

InvalidError · Aug 7, 2014

EuroSong :

Software did not cause your NorthQ PSU to mysteriously fail. Your NorthQ's 5VSB was probably defective to start with and since you have never exercised its higher-power standby role before, enabling sleep mode pushed it over the cliff it was already at.

As for your computer miraculously returning to where it was pre-failure, Windows has "hybrid sleep" enabled by default. With hybrid sleep, whenever you put your computer to "sleep," Windows updates its hibernation file to recover in case power is lost during sleep. If powers does not go out, it resumes from memory but if it does, it resumes from disk which takes several seconds longer.

No power after sleep - bizarre problem

Reputable

Titan

Reputable

Titan

Reputable

Titan

Reputable

Reputable

Titan

Reputable

Titan

Reputable

Titan

Reputable

Titan

Reputable

Titan

Reputable

Titan

Share this page