Probably a dying PSU, but looking for second opinion

[TooNice]

For a while, one of the work desktop was behaving strangely, mostly not switching on using the power button, and requiring flipping the switch on the PSU. Quirky, but since it "worked" I couldn't make a case on doing something about it. Until it suddenly stopped working (while the PC was on) last week. After switching out the power supply (a reputable Antec Truepower; I was half expecting a nasty piece of.. work), everything was working nicely again. Until today. After some testing, I noticed that the PC would turn on if the PSU is switched off for a while (maybe 10min).

So the question I have is, other than the PSU is there anything that might cause this behaviour (e.g. motherboard etc.). Secondly, is this behaviour likely one of a dying PSU, or could this be some kind of failsafe to protect of the PSU* (bit of a wishful thinking really).

* With one dead PSU and another seemingly on the way out, or behaving like the previous one did for a few months before it kicked the bucket, I think that the UPS is likely to blame. I suspected that the quirky behaviour on the last PSU was due to UPS, and a guy I know who work for the company that make the UPS basically call it a piece of excrement that he would strongly not recommend (off the record). On my watch, I probably would've recommended against getting no-brand UPS (much like no brand PSUs), but I am basically the only tech guy around, and when I am not around the guy sent to buy parts will just get the cheapest thing available.

[atemporal]

is wake on LAN/keyboard active? Ie is it truly turning off? If not something might be triggering it to wake. Might also explain why the power button wouldn't turn it on, if it was in a low power state. Though most mobo would also use the powerbutton to wake from too.

It could be PSU, but odd that two brands would behave in same way.

Alternatively Is it a thermal thing? Is something overheating (CPU?) Has the cpu cooler come loose? / a fan burnt out / intake got blocked? What temps are you getting?

If it is the UPS then easy to verify - connect it directly to the mains and see what happens.

[DDY]

I'd put the PSU on the top of the list of things I'd want to check.

If you have access to a multimeter or voltmeter, check the voltage to ground of the purple wire (pin no.9) of the 20/24 ATX motherboard power connector when the machine is plugged in but in a powered-off state. This is the 5V standby rail, which of course should be 5.0V. If it's too low the motherboard won't be able to send the signal to the PSU to fire up when the power button is pushed.

I've had three cases in the past where the machines had similar or exactly the same symptoms as you'd described and all had low or dead 5V standby from the PSU. Waiting for some time allowed the 5V standby voltage to creep up enough for the motherboard to send the power-on signal to the PSU and turn on the machine as normal.

In one case, the problem was exacerbated by a high (but not excessive) current draw from USB peripherals. Unplugging the high power USB devices prior to turning the machine on and re-plugging them once the system had powered-up temporarily fixed the issue, but was permanently fixed by replacing the faulty PSU.

It's possible that a bad output from the UPS could be shortening the life of the PSUs when running on battery backup. I think the explanation for why this maybe is plausible and convincing - there's lots of discussion regarding this on the Internet, but based purely on anecdote I've not experienced any failure of equipment connected to cheap UPS'es.

[TooNice]

@atemporal @DDY

Definitely not wake on LAN/keyboard. Earlier, the desktop shut down while someone was working on it (after a few hours).

It is not two brands, but two PSU of the same brand. However, I believe that the computer was working quite normally prior they changed UPS a few months ago (since then, the PC no longer switches on without first flicking the PSU and now even that won't work before letting the PC "rest" for a while).

I am also starting to wonder if it is not something other than the PSU. What are the odds that a motherboard could be the cause (the horror, if that is the case)?

Heat can't be ruled out. The fans seem to be spinning fine, but the desktop (overbuilt for it's purpose - what's with the i7 just for a simple accounting software) is placed near the ceiling (where heat accumulate) right next to the heat generating UPS.

That said, the fact that the machine can not immediately be started again (as opposed to restart) after shutting down, even though it was only switched on briefly to log into Windows and shutdown in the air-conditioned room I use to test the computer, would suggest that it might not be the fault might not be the heat directly (but perhaps indirectly damaged some parts).

I think that tomorrow, I might do some stress test to see if that causes the machine to shut down by itself (I am not the main user of the computer and haven't actually been able to reproduce that). Are there any diagnostic tools that logs the temp and voltage readings on the fly so that if it does suddenly shut off, I would know what was the last reading?

And about testing the rails with a multimeter, was the 5V rail constantly low, or could it start off normal then suddenly fail, but only after an extended period of time?

[DDY]

To explain in a bit more depth, there are two 5V rails in ATX power supplies, the 'regular' 5V rail - which is indicated by red wires and the 5V standby rail indicated by a purple wire.

The 'regular' 5V will not turn on until the motherboard sends a signal to the PSU to fire up.

The 5V standby rail is always on when the PSU is connected to the mains, this provides the 'background' power to the motherboard and pertinently its power management system, which is what sends the soft power on signal to the PSU when the front power button is pushed.

In the cases I described, it's the 5V standby rail which was problematic, which the 'regular' 5V rail was fine once the PSU powered up.

I didn't observe the 5V standby rail long enough to note any characteristic failure pattern, other than the affected machines would refuse to power up when the 5V standby rail was low or zero. And, trying again after some undefined time later, the 5V standby voltage may have increased allowing the machine to start.

I forgot to mention, that a hard-power cycle, e.g. pulling and re-plugging the mains or turning the PSU switch off and back on again, would occasionally and temporarily push the 5V standby rail voltage back up.

A UPS shouldn't generate much heat unless it's running on battery power, if it's consistently hot while on mains power - which it shouldn't be, I'd consider replacing it.

[TooNice]

So, I finally installed a couple of diagnostic tools. Would've done it sooner but desktop is usually used during my normal working hour and completely offline, meaning that I basically have to get the applications on a USB stick, climb a ladder, and stick it in, in between work hours.

The temperature of the things that I recognise (CPU, GPU etc.) are all fairly normal on idle (low 40s C), so I don't think there is an issue with the cooling.

However, the very instant I tried to run IntelBurnTest, the machine switched off. Instantly. Since I've never used that application before, at first I thought that perhaps it was loading, or the test is done with the screen off, but soon I found out that the desktop is completely off.

I think that I can still rule out temperature as it happened far too quickly. But I can definitely see the unit shutting down as soon as the energy requirement spiked, due to something wrong with the PSU. Still, is there any possibility that it might be due to the motherboard? It does seem weird that it happened to two PSUs within such a short time, unless the last UPS really did something to mess it up.

Anyway, a good replacement PSU is arriving tomorrow. Can you believe that the capital city of the country I am right now doesn't have a PC retailer that sells PSU? You can find graphics card, SSD HD, and a host of component. You can even buy some gaming PCs. But you tell them that you want to buy a PSU? "Sorry, we do not sell those". Like.. What? /Rant

[TooNice]

Well, the problem did go away after replacing the PSU (which I had to ask someone to bring over from abroad).

Thank goodness it wasn't the motherboard.

But considering that PSU 1 worked fine for almost five years (lightly used) and gradually got more quirky the moment they got that dodgy UPS four months ago, and PSU 2, which took the place of PSU 1 after the later started having trouble starting up the PC soon develop the same problem, I think that it is very likely that the now disposed UPS did -something- to drastically shorten the PSU's lifespan. Either that, or a very hard to swallow coincidence..

[DDY]

Does the UPS go in to battery power often?

[TooNice]

The suspect UPS has now been replaced. I don't think that it was switching to battery mode unexpectedly, but we do experience power cuts over once per week on average where it would cut off.

But as far as I know (I wasn't here when they first setup that UPS), shortly or immediately after they set up the UPS, the computer required flipping the power supply switch before you can turn on the PC.

[DDY]

All the cheaper UPS'es I know of just pass the mains straight through unless the voltage or frequency goes out of tolerance, or completely switches off of course, when it then turns to battery power.

So, there really aren't any plausible mechanisms which could explain how a UPS could kill something that's plugged in to it - because as far as the PSU's electronics as concerned the UPS may as well not be there, unless the UPS switches to battery power and outputs some approximation of an AC sine wave. Also, the act of switching from mains to battery (and back again) which can induce spikes in to supply and the momentary loss of power can also be detrimental to the health of a PSU.

If I had to make a bet, I'd put it down to coincidence.

EDIT: I suppose the UPS's own power supply could be defective or badly designed and is putting noise in to supply, or something of that ilk, that's upsetting the PSU?

Still, my money is on coincidence.

[TooNice]

It just seems like quite a coincidence for two respectable and seemingly well functioning PSU to misbehave shortly after plugging into the PSU. I won't rule it out, but I won't take a chance either, especially how hard it is to acquire a simple PSU in this country (whyyy).

The new PSU seems to be working fine with the new UPS, but it is still a bit early celebrate. The new UPS is one that I have now used for a few months without any problem so I think it should be fine. But the guy who bought the PSU decided ignore my recommendation for a well rated Seasonic, and took the salesman's recommendation of a PSU that is "better than Seasonic", yet none of my Googling came with any results. It is cheaper, and capable of more watts than the PSU I suggested and therefore must be better right?

Yeah.. /nervous

[ik9000]

It is cheaper, and capable of more watts than the PSU I suggested and therefore must be better right?

Yeah.. /nervous

When bean counters get to play god with things they don't understand. I have it all the time. Try getting a building built.