PC power up / unreliable startup debugging

[bd139]

So I built myself a PC a few weeks back and it has one annoying persistent problem. It doesn’t power up first time every time. You have to turn it on and off a few times with the soft power button to get it to come up. Once it’s going it’s absolutely rock solid and passes 24 hour memtest86.

Starting this thread as a “debug thread”. If anyone has any suggestions please let me know

Power supply is an Be Quiet Pure Power 11 and the board is an ASRock B450M Pro 4. Nothing is overclocked.

Debugging so far shows correct application of PWR_ON# but an RC curve on the ATX PWR_OK signal which is border line iffy. And it seems to be thermal related (probably crap caps)  as it starts up fine when warm.

So far I have tried:

1. Moving devices to non USB 3 ports. This improved the RC curve. But not enough. This was coinicidental.
2. Checking voltages. All fine when powered up. .
3. Visual inspection of caps. None looking dodgy.

Therefore next steps are:

1. Wait until it’s cold tomorrow morning
2. Blast the board with a hair dryer for a bit to warm it up.
3. See if it comes up. If it does, replace the board.
4. If not, wait until next morning, blast the power supply with a hair dryer and replace that if it helps.

Amy more ideas or thoughts appreciated.

[LateLesley]

Well, start with the basics. Is the switch actually ok? Because many a PC engineer has been caught out with bad case switches. (Me included!)

Second, you say it improved moving USB stuff about. It could be to do with which USB's get powered with what. A PC PSU is actually TWO PSUs, there is an always on one when AC is supplied +5Vsb, and then there is your main one with all the major rails, +12V +5V +3.3V and -5V (I think, or -12V, one was removed in the revised ATX 2.1 spec).
Now, some motherboards will power some USB ports from the +5Vsb, some only power from the +5V, some boards let you choose where some ports are powered from. The +5Vsb supply is limited, it can be anywhere from 1A - 3A (higher on better supplies) but it is possible if you have high draw USB devices to overload the +5Vsb, and in some cases it'll prevent starting cos it sags or trips out on protect. Though you usually need to remove the AC and restart the PSU if it's in protect.

Anyway, its a long winded way of saying, make sure you know which USB ports are being supplied by which power rail (I should be in the manual), and figure out how much the +5Vsb can supply (PSU manual), and how much your USB devices are drawing. (RGB keyboard by any chance?)

Hopefully that gives you a place to start.

[bd139]

Thanks for the reply. Yes the switch is ok. It does actually illuminate the power button when you press it but does not boot the machine. When you press the power button again, the power light goes off. Just no boot action takes place. I've got a backlit keyboard but I did try it with a generic keyboard and that didn't make any difference.

I've pulled all USB devices other than a generic non backlit keyboard and that made no difference.

[LateLesley]

Ah, if the power light comes on, that's a different problem. That means the BIOS isn't booting.

A few things you can check.(well things that can prevent booting) -

Check the bios battery is good. Check the BIOS Battery jumper  (CLR CMOS)
Memory could be an issue preventing boot, so try with one stick in the first slot.
Check your BIOS version, and see if an update is available - in some cases you need to update for some CPUs to be properly supported.
I'm not familiar with UEFI bioses, but I have heard there can be issues there, maybe worth a google. (I think they store drivers in them now??)
And there's the more obvious check all power connectors are tight, no loose ones shorting to the case (though a power light means this isn't the problem)
Maybe conductive thermal paste on the CPU perimeter?
Rogue stand off under the motherboard shorting?
Check the reset switch isn't holding it in reset?
That's about all I can think of, before saying to RMA stuff.

[Specmaster]

Check your power supply, this might help

Edit As LateLesley says, check your reset button has not either stuck or is short circuited, I've know that to happen before too.

[bd139]

I actually didn't connect the reset button because it's on top of the case and I don't want to accidentally press it 

To note the fans all spin quite happily while it's powered up, just the unit doesn't boot!

Watched that video already actually.

I may replace the power supply anyway as I'm not too happy with this one as it's not modular so there's a fat bundle of cables stowed in the bottom of the unit.

[Specmaster]

Are you getting fan spin when it starts up?, is the CPU fan running, if this fails to run, the cpu will overheat very very quickly and shut down even before you get to the normal boot screen.

[Specmaster]

Grab one of these power testers, I have one and they are a very handy tool to have and can help resolve a duff power supply.

[bd139]

I've tested all rail voltages. Absolutely spot on. Not sure I need a specialist tester.

[Specmaster]

Most startup problems are power related.

[Specmaster]

More ideas

[PA0PBZ]

 

[bd139]

Hahaha I know all about those. Interesting fact: my father actually designed and sold the first third party POST card on the market in the early 1980s. All it did was listen to writes on IO port 0x80 and latch it onto the display. It was all discrete 74LS logic. Made a killing out of those. He went to Taiwan and got it produced there and imported a whole container of the things into Europe.  Next thing he knows they were on the clone market for a couple of dollars each less and everyone was selling them .

Our house was a warehouse for about 10 years with thousands of imported IO card stacked up everywhere

[Specmaster]

Yep, I even have a couple variants of those myself, hardly ever used them, infact the last time I relied on of those was so many years ago I have forgotten how to use it even 

Edit I still think its more power supply then anything else, shame you don't have a known good one to swap it out with to prove it either way.

[Black Phoenix]

If it was Intel CPU I would tell you to take out the cooler, and re-sit the CPU and cooler with equal pressure. Some SKT1151 motherboards are finicky with too much pressure on the CPU or uneven one and make stuff like memory channels not being detected and problems with IO and startup.

But since is a Ryzen CPU that doesn't apply, since its PGA instead of LGA.

[ciccio]

I had a similar problem with a number of PCs I have assembled some years ago.
The power button (a large round button on the case's front) in made of some low quality plastic and with time it deforms and it stucks in it's hole, so the power switch  remains pressed, resulting in a power off.
So the PC starts, all fans spin, and it goes off immediately.
The strange thing is that the power button's LED (which is connected to + 5V, not to the motherboard's pins), remains ON.
I used some grease on the button's sides and everything went well.

Best regards

[bd139]

It's not the power button. The right signals are being send/received by the power supply. To some degree that is.

[Black Phoenix]

Well the only thing I will say is if you didn't have the last UEFI update install it. It's not that certainly but is another thing to remove from the equation.

https://www.asrock.com/mb/AMD/B450M%20Pro4/#BIOS

[bd139]

It has 3.50 on it and a Ryzen 5 2600X Pinnacle Ridge. 3.60 release notes say "ASRock do NOT recommend updating this BIOS if Pinnacle, Raven, Summit or Bristol Ridge CPU is being used on your system." so I didn't. 

[Black Phoenix]

It has 3.50 on it and a Ryzen 5 2600X Pinnacle Ridge. 3.60 release notes say "ASRock do NOT recommend updating this BIOS if Pinnacle, Raven, Summit or Bristol Ridge CPU is being used on your system." so I didn't. 

Correct, I didn't know what CPU you had. Yes leave it there, the 3.60 is mainly for Ryzen 3000 and the ABB one is to solve the problem of the Ryzen 3000 not boosting to the advertised speeds (or try to solve because it's still not that linear).

That's really the most strange problem.

[bd139]

Yes indeed it is. And typically I always trigger the strange ones.

I may swap the power supply out first and see if that cures it. I’ll order a different series modular one. I bought this supply because it is extremely efficient and silent. If that doesn’t fix it I’ll swap the board. If that doesn’t fix it I’ll drop kick it out the window and buy a dell 

[Black Phoenix]

Yes indeed it is. And typically I always trigger the strange ones.

I may swap the power supply out first and see if that cures it. I’ll order a different series modular one. I bought this supply because it is extremely efficient and silent. If that doesn’t fix it I’ll swap the board. If that doesn’t fix it I’ll drop kick it out the window and buy a dell 

If you want to trigger even more strange problems that's a good solution

Naa joking, I used extensively Dell PowerEdge R620/R720 in some projects and they are not bad, specially since there a lot of good deals online for used ones. Lenovo ThinkSystems are also kinda reliable...

[/offtopic]

PSU or MB, I point on MB instead of the PSU. If it was the PSU you would find some strange stuff in the voltages when you measure it under load. Other that an intermittent problem that only happens when warm I don't see any other reason.

Ohh by the way, stupid idea but listen to me:  Check the Pins in the ATX 24pin and the CPU 8Pin. Both Motherboard and PSU. Specially PSU, since they can in some cases easily not being totally locked in place and move outside their respective channel.

[bd139]

Hahaha yes. This is actually my first non commercial machine for a while. Usually go with Lenovo for laptops and HP (Z series) for desktops but fancied a custom build.

Makes sense with the PSU vs MB. The PSU is going to be replaced anyway as I bought a non modular one (by accident as I was half asleep when I ordered this) and the Cthulhu nightmare of cables is still lurking in the bottom. So I might as well spend and try and make that go away first. If it happens to solve the issue at the same time then it'll save me the cost of another board. If I end up with the issue after that it'll be a graphics card replacement next. If all that doesn't work then I'll donate the parts to my eldest's PC budget (she's saving up for a gaming PC at the moment)

[bd139]

Ok so current theories are:

1. Power supply is shit.
2. Motherboard is shit.
3. Graphics card is shit.

Easiest option is buy all three and then see which one causes the problem, then return the rest or keep it round for my eldest's PC build. But that means spending £££ which is off the table as I've got a massive tax bill to pay 

So cheap ass diagnostic approaches available are:

1. PC speaker beep (thanks mnem). I will dig out a speaker I have and plug it in and see if it beeps on startup or not.
2. Hair dryer attack (is this thermal)?
3. Power supply tester thing from China [eliminated - I have a DMM]

[magic]

Try less RAM or at least take it out and put back in, maybe in different slots

And what the heck, try it with a POST card.

[bd139]

Done the RAM thing.

Don’t have a POST card. I will try speaker and hairdryer next 

[Specmaster]

You can get a power supply tester delivered tomorrow if you have prime from Amazon for £8.99 https://www.amazon.co.uk/Boladge-Detection-Display-Computer-Supply/dp/B073M2FFJ4/ref=sr_1_4?keywords=pc+power+supply+tester&qid=1572890289&sr=8-4

[Specmaster]

You're using an Asrock MB and I once had an Asrock MB and that was a constant source of trouble as the Bios was unstable and I had to upgrade the Bios several times, working with their support team resolve the problems I was having problems with getting it to boot after upgrading to the then, latest AMD Athlon. I now use Asus MB's and have had zero troubles with them.

[bd139]

Yeah I’m expecting it to be that at the end of the day. I’m trying to take an analytical approach otherwise I just end up spending lots to resolve a minor problem.

I’m not sure what the advantage of the power supply testers are over a DMM. In fact they don’t test all the lines and the resolution is poor and the calibration unknown 

[Specmaster]

Hmm, I think that they do test all the lines otherwise they are pretty pointless. I had problems with a power supply once and despite using a voltmeter, it checked out OK but putting the tester on it it showed that the 3.3v and 5v were low.

[magic]

Hmm, I think that they do test all the lines otherwise they are pretty pointless. I had problems with a power supply once and despite using a voltmeter, it checked out OK but putting the tester on it it showed that the 3.3v and 5v were low.

And how would that miracle work?
If the problem is short term sags or ripple, a scope should be able to catch it.

[soldar]

Well, start with the basics. Is the switch actually ok? Because many a PC engineer has been caught out with bad case switches. (Me included!)

Arghh! Me too! And I never remember when it happens again. Last time I wasted so much time I ended up taking it to the friendly neighborhood computer guy and he found the cause in no time and charged me the minimum of one hour, I think it was 45 EUR, and I remember thinking it was the best spent money because I was testing all the components separately and they were all good. And it was not the first or the second time i had encountered that same problem.

When I built my Win XP system, about 2010, I was having some similar startup problems which were difficult to diagnose. I suspected the power supply but was not sure and did not want to open it and void the warranty so I connected some big-ass capacitors externally in parallel with the output. That seemed to resolve the problem but since the computer was working I never replaced the PSU. Almost ten years later there's still a bunch of capacitors dangling around in that case.

[Zucca]

Replace the switch with a good old one

[soldar]

That switch looks like totally fake to me. Merely for decoration.

[bd139]

Yeah looks like it's made from 1970s butter knives

Haven't had a chance to investigate this further yet

[bd139]

Ok this mystery is solved. I took the PC to bits and jump started the ATX supply on its own and noticed there was an RC time constant on the PWR_OK line as initially. That looked iffy to me so I cracked the power supply open and found the cause! Dry joint on the PWR_OK line. I flooded the joint with some silver bearing solder and it fires up every time fine now.

I suspect this probably passed self test in the factory or wasn’t even tested 

Alas the warranty on the power supply is now voided but at least it works and didn’t cost me anything to fix!

[SiliconWizard]

What's the brand of the PS so we never buy one of those?

[bd139]

Brand is: Be Quiet!  Model is Pure Power 11.

I reckon, as usual, out of millions sold I only get the bad one 

[SiliconWizard]

Brand is: Be Quiet!  Model is Pure Power 11.

I reckon, as usual, out of millions sold I only get the bad one 

Thanks.
Actually those two poor guys seem to have had bad issues with the model 10 (see the two customer reviews): https://www.newegg.com/p/1HU-004H-00065
but judging on the reviews, I can't tell whether this is not just pure trolling.
Let us know if yours ends up exploding.

[bd139]

Ugh I hope that doesn’t happen. Mostly because I have to deal with Aria returns then.

Anyone got any recommendations for a quiet power supply (this PC is silent!) that isn’t a piece of shit? 

[SiliconWizard]

I don't have a lot of experience with Be quiet. They are supposed to make rather good stuff. I for one have been purchasing Enermax PS almost exclusively for ages now (before that I had two PS failures, so I upped my game, Enermax PS being not really cheap, but never regretted it.) The recent models are quiet enough.

[Black Phoenix]

Anyone got any recommendations for a quiet power supply (this PC is silent!) that isn’t a piece of shit? 

Seasonic;
Corsair (some models are manufactured by Seasonic and just have the Corsair branding);
EVGA;
And Be Quiet to... They are not bad PSUs, your case is an isolated one I'm sure.

I had a Corsair AX1200i in one of my PCs and a Be Quiet SFX on a smaller HTPC back home and both are very reliable. Well I had 2 because or shipping error from Corsair, sold one with the PC and kept the other one new, intact on the box.

The best website in terms of Power Supply Units reviews - https://www.jonnyguru.com/blog/category/reviews/power-supplies/

[magic]

silentpcreview was a great resource but it seems they stopped doing new reviews so you would be limited to stuff released a few years ago.

[MrMobodies]

I once had a customer some year ago who brought in a I think it was a Gigabyte motherboard with an I7 cpu, it was new at the time and it had an Antec power supply, it was one of the models where the cables were all black, no colour coding. It would boot up and hang or hang during bootup or not turn on at all.

I tested the voltages with a Dr. Power 2 tester and voltages on all the rails were spot on.

Now when I tried one of my spare power supplies it was fine and no problems, so they took it away, returned their faulty Antec and chose another make and they returned mine back a week later.

That's why I normally keep plenty of spares for this sort of thing.

[bd139]

Yeah definitely worth having a couple floating around.

I just bought another 16Gb of ram for this one. Then it’ll replace my work desktop PC as well which is a somewhat ageing HP Xeon desktop. Then I shall have to buy a backup machine just in case (this pays the bills!) 

[Thecheshirecat]

It’s always the RAM or Motherboard