HTPCs dying inexplicably

[Torajirou]

Hi,

15 months ago, the HTPC that's described in my signature below died on me: it would reboot after a few seconds, again and again. I suspected the picoPSU to be the culprit, tried a standard ATX PSU but the issue persisted. I tried with another RAM module - same issue. So basically it came down to the CPU or motherboard. Since I couldn't easily find parts to pinpoint it and it was all out of warranty anyway, I decided to perform a small upgrade, and bought:

* CPU: Intel i3-4330
* Motherboard: Gigabyte H97N-WIFI

Everything went fine for 15 months, until I came back from holiday yesterday (HTPC was off when I left) and the same problem occurred. I suspect the problem to be identical, though I haven't tried anything yet, really. I'll bring it to the local shop ASAP - I haven't got the time, courage and parts to diagnose it myself.

So my question is as follows: IF my motherboard and/or CPU are burnt again, what could I possibly do to make sure that it doesn't happen again 15 months from now? What could be the cause of these repeating failures and how could I make sure that I am taking the right steps to prevent them from happening again?

Addendum: I just remembered that when my HTPC died the last time, my girlfriend's PC also died on us (same symptoms). Funny enough, it ALSO happened when we were on holiday. Both machines were plugged in (different) Netgear PLC adapters - the only point in common I can think of.

[Rhynri]

Are your machines behind some kind of UPS? If not, that might be part of your problem right there. If you are getting spikes down your power line, that over-voltage and under-voltage can damage electronics. A surge protector strip is generally useless (at least you should consider it as such). A UPS will protect you from all sorts of power line nonsense, and generally line-condition your power to a perfect 120v @ 60hz, reducing heat in your PSU as a pleasant side effect.

One thing that leads me to this conclusion is the fact you have multiple separate machines dying in apparently identical causes at the same time.

[westom]

So my question is as follows: IF my motherboard and/or CPU are burnt again, what could I possibly do to make sure that it doesn't happen again 15 months from now? What could be the cause of these repeating failures and how could I make sure that I am taking the right steps to prevent them from happening again?

You did not identify the defective part before disconnecting or replacing anything. That swapping (shotgunning) has apparently only cured a symptom. The problem remained for another 15 months. Defects can exist and a computer still works. Defective can cause intermittent failures later. That is why a defect is always first identified before replacing anything.

A power controller decides when a system can power off or on. Only it makes those decisions based upon inputs. Shotgunning does not find reasons. Only numbers from a meter can say anything useful.

Other popular urban myths (often seen when shotgunning is an acceptable diagnostic procedure) is that over-voltage and under-voltage damages hardware. Voltage can drop so low that incandescent bulbs dim to below 50% intensity. Even that voltage is ideal for electronics. How often are your bulbs varying intensity that much?

UPS also does not clean power. In fact, 'dirtiest' power comes from a UPS in battery backup mode. Due to hardware inside a computer, even that 'dirty' UPS power is not harmful.

Destructive surges, that can overwhelm superior protection in appliances, is never averted by a power strip protector or UPS. Those devices do not even claim to protect from destructive transients. Only properly earthed 'whole house' protection provides that protection. With numbers that say it does protection. This typically costs about $1 per protected appliance. Even that power strip or UPS needs protection by this proven solution.

[Torajirou]

Thanks for your replies !

You did not identify the defective part before disconnecting or replacing anything. That swapping (shotgunning) has apparently only cured a symptom. The problem remained for another 15 months. Defects can exist and a computer still works. Defective can cause intermittent failures later. That is why a defect is always first identified before replacing anything.

That's my feeling exactly. I realise I cut corners the last time and I don't want to make the same mistake again, because I don't want to fry some new hardware yet again. But what if the defect shows up every other month... Can't it get impossible to identify? Aren't we then stuck with a trial-and-error "shotgunning" approach?

A power controller decides when a system can power off or on. Only it makes those decisions based upon inputs. Shotgunning does not find reasons. Only numbers from a meter can say anything useful.

My problem is that I'm not really sure how to proceed to perform such a thorough examination. Any pointers on how to do that? Anyway, I'm planning to get the whole thing to a technician who might be more competent and better equipped than I am to try and diagnose what could have happened and decide where I go from there.

Other popular urban myths (often seen when shotgunning is an acceptable diagnostic procedure) is that over-voltage and under-voltage damages hardware. Voltage can drop so low that incandescent bulbs dim to below 50% intensity. Even that voltage is ideal for electronics. How often are your bulbs varying intensity that much?

UPS also does not clean power. In fact, 'dirtiest' power comes from a UPS in battery backup mode. Due to hardware inside a computer, even that 'dirty' UPS power is not harmful.

Glad to read that; I suspected so as well.

Destructive surges, that can overwhelm superior protection in appliances, is never averted by a power strip protector or UPS. Those devices do not even claim to protect from destructive transients. Only properly earthed 'whole house' protection provides that protection. With numbers that say it does protection. This typically costs about $1 per protected appliance. Even that power strip or UPS needs protection by this proven solution.

I googled a bit about "whole-house protection" and I got the info that I needed, but I'm still not sure how this could easily be installed - I live in an apartment building (quite a recent one - 2008) and I'm not even sure what's installed already and what is not. Anyway, my options are clearly limited, I guess. By the way, when you say "this typically costs $1 per protected appliance", I'm not sure I understand, considering all the parts and labour involved. Could you please elaborate on that "$1 per appliance" figure?

Anyway, I really doubt that my apartment got struck by lightning or was hit by some destructive transient targeted specifically at my HTPC... twice in 15 months; it is connected to the same power strip as my NAS, my TV, my loudspeakers and a network switch. OK, my girlfriend's PC died apparently at the same time, but that could also be a coincidence (I guess we'll never know).

So, until I get more accurate answers from that technician (provided I ever do get those), I'm still considering what I could do next, still in an "empirical shotgunning approach":

* Replace my current power strip with a surge-protected one (that won't probably hurt anyway)
* Trade my picoPSU for a new one, with a new brick
* ... ?

Thanks again for your feedback!

EDIT: @westom: I'm already reading this thread BTW http://www.techspot.com/community/topic ... un.165850/

[Vicotnik]

What brick are you using with the picoPSU?

[Torajirou]

What brick are you using with the picoPSU?

It's a 150W Fortron brick. More or less the one listed here, I guess: http://www.myelectronics.nl/contents/nl/d348.html (at least that's where I bought it back then - not sure it's still the same part exactly).

[Vicotnik]

Not crap then. Hard to imagine the brick being responsible for damaging the system. Are the voltage levels ok? No damaged caps on the motherboard?

[Torajirou]

Not crap then. Hard to imagine the brick being responsible for damaging the system. Are the voltage levels ok? No damaged caps on the motherboard?

None that I could see, I'll get a closer look when I get home.

I also spent more time reading westom's replies on the thread I linked above and will definitely spend some time learning how to use a multimeter before I seek "professional help"

I'll keep you posted

[Torajirou]

So I bought a multimeter and tried to follow westom's instructions as found there : https://forums.techguy.org/threads/cust ... st-7547787

I had to search a bit because there are no wires on the PicoPSU; found that : https://superuser.com/questions/905705/ ... pec-pinout

Unfortunately, probing some voltages on my setup without at least removing the CPU heatskink is downright impossible - all pins from 1 to 12 are inaccessible... Tell me if it's safe to test without the heatskink, knowing it only stays on for a few seconds...

What I could gather so far:

Code: Select all

                  OFF       ON
13 (Orange)    +00.00V   +03.34V
14 (Blue)      -00.00V   -10.88V
16 (Green)     +04.96V   +00.01V
21 (Red)       +00.00V   +05.06V
22 (Red)       +00.00V   +05.06V
23 (Red)       +00.00V   +05.06V

Any suggestion forward ? :s

[Vicotnik]

Try measure the +12V rail in various places. You can do that from a molex (between yellow and one of the black wires). Or on the 4pin cable connected to the motherboard. You can also try measuring on the brick, without it connected to anything.

Of the points you have measured, what sticks out is -10.88V on pin 14. That should be -12V ± 10%.

[westom]

Any suggestion forward ?

Try using a paper clip to get into those other connections. If necessary, partially insulate that paper clip with scotch tape.

Removing a heatsink is not recommended since good diagnostic procedure must collect all facts before making any changes. If the CPU is not an Intel type, then bad things might happen without a heatsink.

Not only measure voltages. Behavior during power up is key on some readings.

Blue wire (-10.88) is a bad voltage. However that voltage often does not have a function; typically does not affect a computer's operation. In fact, since every voltage should have some sort of load and since -12V typically does not, then a power supply fan is often powered from that -12 volts - just to create a load.

Necessary are voltages from a purple and gray wires - especially behavior of that gray wire during power up.

So far, no relevant problems are observed.

Measurements without a load can often result in a bad supply acting good. A load is essential to measure a defect.

[Torajirou]

OK, so I managed to get some more readings using a paper clip:

Code: Select all

                  OFF       ON
08 (Gray)      +00.00V   +05.05V
09 (Purple)    +05.08V   +05.08V
10 (Yellow)    +00.00V   +12.16V  
12 (Orange)    +00.00V   +03.34V

As for the transitions from the OFF to the ON state that I posted here and above, I notice nothing spectacular; it just goes from one value to the other in less than a second. From ON to OFF, going back to zero usually takes longer, but at this point, the machine is stuck in an endless on/off loop so it difficult to measure.

[Vicotnik]

Nothing wrong there. +12V looks good.

[westom]

I notice nothing spectacular; it just goes from one value to the other in less than a second. From ON to OFF, going back to zero usually takes longer, but at this point, the machine is stuck in an endless on/off loop so it difficult to measure.

OK. The many components of a power system are OK. Some background. Purple wire is power for the power controller. That must be always on (and is why a power cord must be disconnected from a wall receptacle before making any changes). Those numbers also report that a standby power supply (purple wire) is properly isolated from the other 5V (red wire) power supply.

Green wire is a power controller telling a PSU to power on. It is doing just that with numbers that are more than sufficient (and that also say you were measuring correct wires). Gray wire says a power supply has confirming what you also reported - that critical voltages power are stable.

So the power system is just fine. Power controller would be powering off a PSU due to some other input.

For example, if the front panel power button is stuck or sticking, that could be one defect. Measuring that switch for conductivity when pressed and open circuit when not pressed means using the digital meter's 20 VDC volt setting. This may be mechanically hard. Maybe use a sewing needle to touch wires that connect that switch at a motherboard header (connector). When switch is not pressed, it should measure well above 2.4 volts. When switch is pressed, it should measure well below 0.7 volts. If that switch is stuck or sometimes sticks, that can be one input that causes a computer to power cycle.

Temperature is another input. If a controller is told that temperature is excessive, then it may power off. Unfortunately that function can be unique with different motherboard. For example, many motherboards based in Intel CPUs will simply keep working; albeit slower. Some have other behaviors if a fan does not spin.

One rare reason for power cycling can be a dying CMOS battery. It can be removed and measured with the digital meter. If that 3 volt lithium cell is at 2.9 volts, then battery is just fine but probably needs replacement in the next year. Battery voltage would have to be well below 2.8 volts to be problematic. Normally a low battery, by itself, would not cause problems. But that low voltage combined with some other anomaly could explain unstable operation. Easy is to just measure that battery - to eliminate it from a list of suspects.

I am assuming your computer does not even get to the BIOS before shutting off. If it does, well, how far does it get? That may add other suspects (inputs to the controller).

We know your power system (which includes PSU) is OK. That eliminates many suspects. Something else is telling the controller to turn off power. 'When' it is inputting that request can help identify what and from where that request is coming from.

[xan_user]

@ OP:
my bet is still on lack of ups. the nas may have been lucky, had a better PSU or just on a more robust circuit....who knows.
play it safe. Its good insurance to have a ups for every "important" PC.

[CA_Steve]

Another simple thing to test are the AC outlets themselves to see if they are wired correctly and there's not a swapped neutral and hot wire in one. It's amazing what gets past inspection. Find a local version of this.

[westom]

Another simple thing to test are the AC outlets themselves to see if they are wired correctly and there's not a swapped neutral and hot wire in one.

If reversed polarity caused a problem, then numbers from the meter clearly identified that defect. It was not identified by numbers. All appliances work just fine with reversed polarity.

Polarity is for protecting humans. One of many things done to protect human life. Polarity does not affect the operation of any appliance. Since AC is power is even constantly reversing polarity. Other AC line anomalies such as excessive harmonics, low voltage, and open neutral also do not exist according to numbers from a meter.

UPS is another example of a recommendation without basic electrical knowledge. What 'at risk' part would be protected? UPS does not claim to protect hardware. If it did something useful, then numbers are posted to define what it does.

'Dirtiest' power seen by a computer comes from a UPS in battery backup mode. A UPS might be used to make power so 'dirty' so as to identify a computer defect. But UPS does nothing to protect hardware. Otherwise a spec number defines that protection. Otherwise a potentially damaged internal part could be identified.

Numbers from a meter make obvious that AC power is perfectly good. His defect would be something else inside and well protected from AC mains by a perfectly good and robust power 'system'.

[Torajirou]

I managed to get readings from the power button header using a sewing needle and from the battery.

The power button header is 3.39V unpressed and 0.00V when pressed.

The battery is 3.20V.

Please find below a link to the video I just took - the power LED on the bottom right corner is NOT attached to the motherboard header but to a regular 4-pin molex. Notice how sometimes only one fan spins up and sometimes both do... Also notice the sound of the Bluray drive when it starts. Not sure if that helps, but it certainly won't hurt.

https://www.youtube.com/watch?v=OQYbA4YhyhU

To answer one of your previous questions, depending on the time it takes to reboot, I sometimes see some kind of BIOS splash screen, but very briefly. At least, *something* is output on the screen...

(Sorry about the noises made by my girlfriend cleaning the windows in background )

[CA_Steve]

Two separate issues. One is debugging his current PC. Second is when three PCs die, two at the same time it implies a systematic problem with his mains / power quality issue.

Using the $5 outlet tester is a simple way to see if, at least, the apartment is wired correctly.

Torajirou - Both PCs are plugged into Netgear powerline networking adapters...how well are the Netgear devices working? Also, according to Netgear product FAQs you can't have a surge protector or UPS in line with a PLC.

[Torajirou]

Two separate issues. One is debugging his current PC. Second is when three PCs die, two at the same time it implies a systematic problem with his mains / power quality issue.

"at the same time" may not be 100% accurate; rather during the same time frame (a couple of weeks or so)

Torajirou - Both PCs are plugged into Netgear powerline networking adapters...how well are the Netgear devices working? Also, according to Netgear product FAQs you can't have a surge protector or UPS in line with a PLC.

I do have power strips plugged into both. One of them is supposed to be a "surge protector", but it's very old and cheap stuff. The HTPC is plugged into a regular power strip. The network speed is very fast and completely stable. From what I read in your link, Netgear does not seem to warn against my setup (a power strip plugged into the PLC adapter), rather against the opposite (a PLC adapter plugged into a power strip).

[CA_Steve]

Yeah - should have been more specific - you can't have surge protection between your PLC nodes and the wall socket.

[westom]

The power button header is 3.39V unpressed and 0.00V when pressed.

That 3.39 number is troubling. It should be higher. Normally a pullup resistor would make it closer to 5 volts. At 3.39, that might only be leakage currents from the power controller. So that noise could easily cause the controller to think the button is pressed.

Try this. Get a jumper cable (ie alligator clips) and a resistor (ie from Radio Shack) that is anywhere between 1K ohms to 100K ohms. Connect that resistor from 5 volts (ie any red wire) to the header pin that measures 3.39 volts. The computer should power on and stay on - stop cycling - if voltage is closer to 5 volts. Then we know a pullup resistor is missing or failed.

Left fan is apparently not reliable. Fans need maximum torque on startup. Sometimes a Hall Effect sensor inside a fan is not properly seated or bearing grease has become sticky. Then a fan does not start every time. Flick that fan with a finger and it would run constantly. That means the fan is defective.

Fan probably is unrelated to your symptoms. But that one fan acts as if it is defective.

[Torajirou]

Thanks ! I will definitely do that and keep you posted !

[Torajirou]

So I went and bought a 10kΩ resistor and some alligator clips and did as you said. The voltage on the switch header is now above 4.10V when the PC is running, still, it reboots after a few seconds over and over again

Any more ideas ?

[westom]

Any more ideas ?

I am out of ideas that can be performed at a layman's level.

To be clear, the green wire should rise above 2.4 volts just before (almost instantly before) a PSU powers off. That is the power controller ordering a PSU to powering off.

Maybe the gray wire is asking for power off by dropping to below 0.7 volts just before. We currently assume the gray wire is dropping to below 0.7 volts AFTER the power supply powers off. Unfortunately that cannot be seen by a meter since it happens too quickly. We assume the gray wire does not request power off because voltages on the red, orange, and yellow wires are so good and stable.

Well the blue -12 volt wire is bad (-10.88). Maybe that voltage is monitored by the PSU and reported on the gray wire. It typically is not. Is there anything that might be powered by -12 in your hardware? -12 exists only for special analog devices that most computers do not have. Is there any unique peripheral card in one PCI (peripheral) slot?

Again, we assume the gray wire is not requesting a power off. Unfortunately to actually confirm that would require something like an oscilloscope. However does anything exist that might be overloading the blue wire - cause -12 volts to be insufficient?

Other functions that can cause a power controller to power off a PSU tend to be unique to a manufacturer's design. You say something does flash on the screen (most times or every time?). That comes from the BIOS.

BIOS must execute a complex set of instructions to setup the PCI bus, memory configurations, and other peripherals. When it assigns a simple video mode on that controller, it then flashes up a text message in the upper left corner. This occurs long before it even looks at any disk drive or tries to boot Windows. BIOS normally does not / cannot request the power controller to power off. So BIOS typically is not in a suspect list of functions that would request a power controller to power off. If it was, the BIOS would get to the same point every time that power off occurs.

IOW I am searching for functions that would tell a power controller to power off. That can be observed using layman tools. And that are standard across all motherboard designs.

This reviews what we do and do not know hoping something was overlooked or not accurately observed. Power controller (located in large adjacent ICs also known as the CPU chipset) is ordering the PSU to power off. Why (what input is requesting that power off) is not apparent. Even the purple wire voltage (a more common source of unstable voltage (excessive ripple) and therefore unstable power controller operation) measures just fine.

If a front panel push button switch is stuck at too low a voltage (ie stuck closed), then even that would request the power controller to power off. Duplicate that defect by holding the power button down constantly. But your numbers indicate that power button is not closed (stuck) when not pressed. Your numbers even exonerate that suspect.

A fan that is suppose to be spinning and does not might be one input - unique to a motherboard's design. But power off occurs both when that one fan does and does not spin. I am assuming those fans are original - therefore have a sensor wire that reports the fan is spinning (if required) so that the power controller knows those fans are spinning when they are suppose to. Better motherboards even provide software that can report how fast each fan is spinning. And temperature sensors that can report what those temperatures are and that can slow a fan when temperatures are low enough. Those are also BIOS functions that might request a power off when hardware is not working correctly. But the video suggests that is not happening with each power off.

Does this motherboard have any other unique hardware that might request a power controller to cause a sudden power off? I have run out of usual suspects that can also be observed with simple equipment, without a motherboard schematic, and without knowing of unique functions / inputs to that power controller.

Again, just to confirm - power is cycling off after and because that green wire voltage rises to above 2.4 volts. That is a power controller telling a PSU to power off. We have run out of usual suspects that are requesting that power off. Above is a summary of those usual suspects and why each one is exonerated.

[Torajirou]

Thank you so much for the time you devoted to my issue.

I learned a lot, even though it didn't turn up for the best

[Vicotnik]

If it's not bad power, then maybe overheating? Or the motherboard don't get a RPM reading on the CPU fan and shuts down? Grasping at straws here..