Help out a noob at motherboard repair

[LeonR]

So, I've started a thread on badcaps (dunno if I can link it) about changing caps, but as some users posted the problem of one of those boards (the Tyan) may not lie on the caps themselves. The board is a Tyan S7012 (dual 1366 server-grade).

The symptoms the board is showing are the following:

* When I try to use 2x X5690 (130W TDP), the board power gets cut while in POST and it refuses to power up after it. I need to disconnect power, force a "power flush" (pressing the power on while PSU is disconnected) just for the whole thing to loop.

* I tested 2x E5649 (80W TDP). Installed Windows, did a lot of RAM testing (bootable memtest86+) and it work just fine on those basic tasks. When I tried to stress test the system with Prime95, I started hearing some "fshhh" noises, that remembered me of caps venting - but none of them are actually vented out on the top. I noticed that the mouse would freeze for a second or so sometimes, and the noise stopped while on this state. It didn't take long for the board to freeze completely (about 20 secs) and the noise to stop. Weird thing is that the board worked fine with those CPUs yesterday, even on the stress test.

* I remember testing the board with 2x 5504 (80W TDP but probably runs o a pretty low power) and the sizzling noise would be present while stress testing, but the board did hold up for the whole test.

* When I got this board I had some Intel coolers with a low bracket that ended up lifting one of the VRM's coils on one side. I turned the board on since it was unknown to me at the time - I found out about this a day later when I was swapping CPUs to test. I've resoldered it when I found out, but maybe some of the VRM ICs got toast? There isn't any smell, PCB discoloration or clear physical damage on them.

I'm using a EVGA 750BQ which is pretty new and was used to tested another server board with 2x X5690 and fully populated RAM, no problems at all.

I own a multimeter (Aneng 8009), soldering tools, a flash programmer (TL866II Plus) and lots of spare parts to test.

Could you please try to help me out? I'm kinda lost on where should I start.

[janoc]

If the board boots with another CPU then it is not very likely the problem is in the capacitors. Blindly replacing anything, be it capacitors or something else is never a good idea - all you could achieve is that instead of only having the original problem you may now have lifted traces, scorched board and some cooked parts in addition to it. And in the best case you have only wasted money on the caps.

What you are describing sounds very much like the board shutting down because of some emergency - some voltage regulator overload due to excessive current, the board not sensing spinning fans/sensing the CPU heating up too rapidly, etc.  All that will trigger an alert and instant shutdown in order to protect the machine.

The noises you have heard and the instability you had with the slower CPU seem to point to some overloaded voltage regulator. It barely works with the weaker CPU and the board shuts down with the more power hungry one because it is drawing too much current.

The sizzling could have simply been something becoming too hot or the inductor of the DC-DC converter sometimes makes such noise when under high load, not a capacitor venting (that you certainly would not miss - it stinks and there is smoke!). Moreover, an electrolytic cap will vent only if it is suddenly overheated (either because of a fault or a soldering iron) or someone installed it backwards. However, if it was working for a long time, it may be dry or have leaked but it won't suddenly vent - that needs boiling electrolyte and a dried out cap won't have much of it left.

Are you overclocking it, by chance? Any non-standard BIOS settings? Fans/heatsinks properly installed? I had a machine  crash on boot like this because one screw/prong of the heatsink fixation system was loose.

What about the power supply? These server class boards need beefy supplies and if the CPU rail (you do have those CPU power cables plugged in, right?) isn't able to deliver enough current because of a cheap/crappy power supply or not all of the cables plugged in, you would see these symptoms too.

[LeonR]

If the board boots with another CPU then it is not very likely the problem is in the capacitors. Blindly replacing anything, be it capacitors or something else is never a good idea - all you could achieve is that instead of only having the original problem you may now have lifted traces, scorched board and some cooked parts in addition to it. And in the best case you have only wasted money on the caps.

What you are describing sounds very much like the board shutting down because of some emergency - some voltage regulator overload due to excessive current, the board not sensing spinning fans/sensing the CPU heating up too rapidly, etc.  All that will trigger an alert and instant shutdown in order to protect the machine.

The noises you have heard and the instability you had with the slower CPU seem to point to some overloaded voltage regulator. It barely works with the weaker CPU and the board shuts down with the more power hungry one because it is drawing too much current.

I forgot to mention: I've tested the board yesterday (stress test) with the 2x E5649 and it just worked fine.

The sizzling could have simply been something becoming too hot or the inductor of the DC-DC converter sometimes makes such noise when under high load, not a capacitor venting (that you certainly would not miss - it stinks and there is smoke!). Moreover, an electrolytic cap will vent only if it is suddenly overheated (either because of a fault or a soldering iron) or someone installed it backwards. However, if it was working for a long time, it may be dry or have leaked but it won't suddenly vent - that needs boiling electrolyte and a dried out cap won't have much of it left.

Nice to know. I had some qualms about the board having heaps of NCC KZGs, which have falied on another board of mine (an Asus A8V - all of them have bulged).

Are you overclocking it, by chance? Any non-standard BIOS settings? Fans/heatsinks properly installed? I had a machine  crash on boot like this because one screw/prong of the heatsink fixation system was loose.

Nope. (most) Server-grade motherboards are suited for stable performance (they are meant to be run 24x7x265) and don't even allow you to go one number above the maximum multiplier allowed by that CPU. BIOS on those is pretty barebones compared to high-end desktop, what you get is more enterprise-level features to set up, like OOB, BMC, error logging, etc. I'm using server-grade heatsink fans on them, and I also monitor temps while testing on Windows (via software).

What about the power supply? These server class boards need beefy supplies and if the CPU rail (you do have those CPU power cables plugged in, right?) isn't able to deliver enough current because of a cheap/crappy power supply or not all of the cables plugged in, you would see these symptoms too.

It is a 750W EVGA PSU with two EPS (the board need two of them). I'm using a grounded and correctly polarized outlet. Everything power concerned is OK, I guess.

-------------------

One other thing I forgot to mention: When I got this board I had some Intel coolers with a low bracket that ended up lifting one of the VRM's coils on one side. I turned the board on since it was unknown to me at the time - I found out about this a day later when I was swapping CPUs to test. Maybe some of the VRM ICs got toast? There isn't any smell, PCB discoloration or clear physical damage on them.

-------------------
I've attached a pic and identified the caps:

NCC KZG (red circle):
18x 470uF @ 16V

NCC KY (yellow circle):
4X 220UF @ 16V

SUNCON (blue circle):
5x 1000uF @ 6.3V

[Johnny10]

I have Dell, Supermicro, Tyan 1366 boards and some won't run with high wattage CPU have you checked the manual for there may be a wattage limit?

X5690's draw a lot of power.
Sure you haven't damaged CPU with Stress test?
I believe some of the benchmark tests state they may damage CPU.
Are you using Intel IPDT (Intel Processor Diagnostic Tool) this tool will show if processor is damaged or overheating.

[LeonR]

I have Dell, Supermicro, Tyan 1366 boards and some won't run with high wattage CPU have you checked the manual for there may be a wattage limit?

X5690's draw a lot of power.
Sure you haven't damaged CPU with Stress test?
I believe some of the benchmark tests state they may damage CPU.
Are you using Intel IPDT (Intel Processor Diagnostic Tool) this tool will show if processor is overheating.

Yes, it does support the X5690. Nope, the board didn't kill any CPUs.

When I stress test the system I keep an eye at the temps with CoreTemp. Also, I always keep a 60mm fan blowing fresh air on the PCH. Those 5500/5520 get VERY hot really quick if left with the passive cooling.

[David Hess]

It is a 750W EVGA PSU with two EPS (the board need two of them). I'm using a grounded and correctly polarized outlet. Everything power concerned is OK, I guess.

I have had motherboard problems like these before with marginal power supplies which were either poor quality or had worn out capacitors.  Swapping the power supply is usually the simplest way to test this but if I wanted to know for sure, I would monitor the power supply output voltages with an oscilloscope with various CPU loads.

[Johnny10]

One more possibility if you have eliminated Power Supply as problem.

I have Supermicro dual 1366 motherboards that state they will run X5675( 95Watts) x-5690( 130Watts) but the Bios needs to be upgraded.
SuperMicro website specifically stated would work with original BIOS but without upgrade would not.
If you can get the motherboard to POST with E5649 (80watts per CPU) then nothing wrong with motherboard.

I currently run Tyan S7025 1366 Boards with 2-x5675 and triple redundant 800W Power Supplies.

[janoc]

One other thing I forgot to mention: When I got this board I had some Intel coolers with a low bracket that ended up lifting one of the VRM's coils on one side. I turned the board on since it was unknown to me at the time - I found out about this a day later when I was swapping CPUs to test. Maybe some of the VRM ICs got toast? There isn't any smell, PCB discoloration or clear physical damage on them.

If the coil hasn't been torn off the board, the winding physically damaged or some traces on the board damaged then it is likely fine. Merely moving it won't damage anything.

-------------------
I've attached a pic and identified the caps:

NCC KZG (red circle):
18x 470uF @ 16V

NCC KY (yellow circle):
4X 220UF @ 16V

SUNCON (blue circle):
5x 1000uF @ 6.3V

Then get an ESR meter (not a capacitance tester) and measure them in circuit. That's the only way to know whether they are good or bad. Nobody here can diagnose that from the photos.

However, given the symptoms you have, I would say you have a power supply problem and there is nothing wrong with these caps - if they were bad then even the weaker CPUs wouldn't boot in it.

[LeonR]

It is a 750W EVGA PSU with two EPS (the board need two of them). I'm using a grounded and correctly polarized outlet. Everything power concerned is OK, I guess.

I have had motherboard problems like these before with marginal power supplies which were either poor quality or had worn out capacitors.  Swapping the power supply is usually the simplest way to test this but if I wanted to know for sure, I would monitor the power supply output voltages with an oscilloscope with various CPU loads.

I monitored the voltages with my DMM, and the PSU don't go stray towards the ATX regulations (10%). Unfortunately I don't own an oscilloscpe.

One other thing I forgot to mention: When I got this board I had some Intel coolers with a low bracket that ended up lifting one of the VRM's coils on one side. I turned the board on since it was unknown to me at the time - I found out about this a day later when I was swapping CPUs to test. Maybe some of the VRM ICs got toast? There isn't any smell, PCB discoloration or clear physical damage on them.

If the coil hasn't been torn off the board, the winding physically damaged or some traces on the board damaged then it is likely fine. Merely moving it won't damage anything.

-------------------
I've attached a pic and identified the caps:

NCC KZG (red circle):
18x 470uF @ 16V

NCC KY (yellow circle):
4X 220UF @ 16V

SUNCON (blue circle):
5x 1000uF @ 6.3V

Then get an ESR meter (not a capacitance tester) and measure them in circuit. That's the only way to know whether they are good or bad. Nobody here can diagnose that from the photos.

However, given the symptoms you have, I would say you have a power supply problem and there is nothing wrong with these caps - if they were bad then even the weaker CPUs wouldn't boot in it.

The inductor is one of the sealed types, SMT. I resoldered the lift leg after I found out, but the board ran for a while (some hours) with it lifted.

I've bought one of those LCR multitesters (haven't received it yet). I assume those can't be used for in-circuit measuring?

[janoc]

I monitored the voltages with my DMM, and the PSU don't go stray towards the ATX regulations (10%). Unfortunately I
don't own an oscilloscpe.

Without a scope you are in the dark - the supply could well be "dirty" because the filter capacitors in it are dead and this a well known source of problems, especially when under load.

The inductor is one of the sealed types, SMT. I resoldered the lift leg after I found out, but the board ran for a while (some hours) with it lifted.

Are you sure about that repair? Because an inductor in a DC-DC converter could be only one of two things - either the main inductor of the converter or a filter choke. In either case, if it was disconnected/torn off then that voltage rail would have been dead.

If this has worked with the inductor damaged and stopped working after fixing it then this would be very much the first thing I would look into - you may have shorted something on the motherboard with the solder or connected the inductor to a wrong place.

I've bought one of those LCR multitesters (haven't received it yet). I assume those can't be used for in-circuit measuring?

You mean one of those cheap Chinese things that tests also transistors and what not? No, that cannot be used in circuit (and don't even try or you could fry something on the board - plenty of stuff there that isn't designed to survive 4.5-5V that these testers put on the pins).

You need a proper ESR meter, something like this:
https://www.ebay.fr/itm/MESR-100-Auto-Ranging-ESR-Low-Ohm-Circuit-Capacitor-Meter-Tester/162850252754

(hint - if it has more than 2 input socket, then it is not an ESR meter but again one of those multi-testers. They can measure ESR but only out of circuit)

[LeonR]

I monitored the voltages with my DMM, and the PSU don't go stray towards the ATX regulations (10%). Unfortunately I
don't own an oscilloscpe.

Without a scope you are in the dark - the supply could well be "dirty" because the filter capacitors in it are dead and this a well known source of problems, especially when under load.

The inductor is one of the sealed types, SMT. I resoldered the lift leg after I found out, but the board ran for a while (some hours) with it lifted.

Are you sure about that repair? Because an inductor in a DC-DC converter could be only one of two things - either the main inductor of the converter or a filter choke. In either case, if it was disconnected/torn off then that voltage rail would have been dead.

If this has worked with the inductor damaged and stopped working after fixing it then this would be very much the first thing I would look into - you may have shorted something on the motherboard with the solder or connected the inductor to a wrong place.

I've bought one of those LCR multitesters (haven't received it yet). I assume those can't be used for in-circuit measuring?

You mean one of those cheap Chinese things that tests also transistors and what not? No, that cannot be used in circuit (and don't even try or you could fry something on the board - plenty of stuff there that isn't designed to survive 4.5-5V that these testers put on the pins).

You need a proper ESR meter, something like this:
https://www.ebay.fr/itm/MESR-100-Auto-Ranging-ESR-Low-Ohm-Circuit-Capacitor-Meter-Tester/162850252754

(hint - if it has more than 2 input socket, then it is not an ESR meter but again one of those multi-testers. They can measure ESR but only out of circuit)

It was an inductor from one of the CPU's VRM circuit. I marked it in red (the picture is slightly different since that is the higher end version of the same mobo).

I've bought a new PSU for another system and I'm going to be able to re-test the board with a different PSU, just in case.

[David Hess]

It is a 750W EVGA PSU with two EPS (the board need two of them). I'm using a grounded and correctly polarized outlet. Everything power concerned is OK, I guess.

I have had motherboard problems like these before with marginal power supplies which were either poor quality or had worn out capacitors.  Swapping the power supply is usually the simplest way to test this but if I wanted to know for sure, I would monitor the power supply output voltages with an oscilloscope with various CPU loads.

I monitored the voltages with my DMM, and the PSU don't go stray towards the ATX regulations (10%). Unfortunately I don't own an oscilloscpe.

Without a scope you are in the dark - the supply could well be "dirty" because the filter capacitors in it are dead and this a well known source of problems, especially when under load.

The last several times for me when mysterious reboots were the problem just changing the power supply was enough to solve the problem and I did not test further and instead examined the power supply itself which always revealed worn out output capacitors.

An earlier incident with a Pentium 2.4C and an Intel 875P chipset was more interesting.  First the built in Intel CSA Gigabit Ethernet failed so I switched to using a PCI network adapter.  Then several months later, the Radeon 9500 Pro started having graphics corruption issues so I fiddled with the GPU and memory clock speeds.  Then finally several months after that, the system had trouble booting if it had not warmed up which was the indication of a capacitor problem; capacitor value and ESR changes with temperature.  The problem the whole time was the power supply and replacing it fixed everything including the built in Intel CSA Gigabit Ethernet which I had assumed was dead.

The time before that, I actually did end up hooking up a DSO to monitor the supply voltages and caught an infrequent glitch that took hours to manifest.

Usually when testing with any oscilloscope, a DSO is not required for this, a sure sign of trouble is excessive high frequency ripple which is not something a multimeter is going to reveal.  The output voltages can be dead on with good regulation but if there is 2 volts of ripple on the +12 volt supply, something is wrong.

[LeonR]

I'd be more inclined to suspect a PSU problem if it was old (it has a bit more than a year) or if I had the same problem with testing another board (I used the same PSU at a more hungry system from the same class), but that isn't the case.

[janoc]

I'd be more inclined to suspect a PSU problem if it was old (it has a bit more than a year) or if I had the same problem with testing another board (I used the same PSU at a more hungry from the same class), but that isn't the case.

But you did have it working with a less power hungry CPU and not working with one requiring more power. That's pretty much a textbook example for some power supply issue where the PSU has problems under load. Either not able to deliver the necessary current or having excessive ripple under load (e.g. because of failed capacitors).

That a supply is "new" (less than a year) doesn't mean much - you don't know how long it was sitting somewhere or how old the capacitors in it are. And if the manufacturer used cheap caps (common even in premium brands), they can easily fail even after few months of sustained use.

[LeonR]

I'd be more inclined to suspect a PSU problem if it was old (it has a bit more than a year) or if I had the same problem with testing another board (I used the same PSU at a more hungry from the same class), but that isn't the case.

But you did have it working with a less power hungry CPU and not working with one requiring more power. That's pretty much a textbook example for some power supply issue where the PSU has problems under load. Either not able to deliver the necessary current or having excessive ripple under load (e.g. because of failed capacitors).

That a supply is "new" (less than a year) doesn't mean much - you don't know how long it was sitting somewhere or how old the capacitors in it are. And if the manufacturer used cheap caps (common even in premium brands), they can easily fail even after few months of sustained use.

I used the PSU to set up another system (an Intel S5520HC) with 2x X5690 and 12x RAM slots populated. That is pretty much on the same power level as the Tyan.

[David Hess]

I usually try to start with the simple to test things like replacing the power supply or measuring the output noise instead of probing the motherboard itself because even if I have reason to think they are less likely to cause the problem, they are so much easier to rule out.  Testing the motherboard itself is a drop down the rabbit hole.

[LeonR]

Off-topic: Last time I've used a oscilloscope was something ~15 years ago.  What would be a good option for PC repair (motherboard, PSU, GPUs, etc)?

[David Hess]

Most of the stuff you can fix involves power which almost any oscilloscope is suited for if it is 20 MHz or faster.  Maybe you can find something used locally to practice with.

[janoc]

Off-topic: Last time I've used a oscilloscope was something ~15 years ago.  What would be a good option for PC repair (motherboard, PSU, GPUs, etc)?

Basically even an old analog scope would do. There isn't much you can do with a cheap scope on a modern PC motherboard or GPU apart from checking power rails. Everything else is way too high speed/too many signals (or both) and thus needs very expensive equipment. Worse, with no documentation for the board even that would be pretty much useless in many cases.

However, if you want to poke around ATX power supplies, do make sure to get a differential probe and an isolation transformer too. OTOH, unless you are sure what you are doing, it is probably best to stay away from those things - it is way too easy to get a potentially lethal shock there. Also there isn't much point in repairing those most of the time, given how much a new one costs - your time and the replacement parts will invariably cost more than a new supply unless it is some super expensive one (and/or a trivial fault that takes 5 minutes to fix - most aren't).

And I don't want to rain on your parade but don't hold your hopes high for being able to repair anything of significance on modern motherboards or GPUs.  Even if you manage to diagnose the fault correctly (except for trivial and obvious faults it is difficult to impossible, given that there is no public documentation on how these things work available), you will invariably hit a wall because the replacement components are simply not available (such as a new northbridge or a GPU) and their replacement requires special equipment (BGA rework station)/is economically unfeasible (things like replacing damaged CPU sockets or swapping GPU chips).

If you look at Louis Rossmann's videos, I believe he explained few times why he is working only on Apple hardware - it is "standard" (in the sense that there only so many variants of it), it is popular and ubiquitous enough so that there are at least some sources for replacement parts and documentation (= grey/black market). And finally the hw is sufficiently expensive so that a client won't balk at a few hundred dollar repair bill for the tech's time. You can't ask $400 for spending a day troubleshooting and fixing a $300 motherboard, nobody will pay that. A $2000 laptop is a possibly a different story.

[LeonR]

Off: I've bought some tools for tinkering with basic stuff - capacitor replacement, BIOS chip removal and reprogramming, etc. since I'm a bit of a collector of x86 hardware and some of them are already showing the signs of aging (got five other boards that need recapping). I'm yet to receive my hot air gun and soldering stuff - flux, wires, etc.

I sent this same board (and a Supermicro) to a repair shop for a cost estimate; dude told me both would cost ~USD500 to get repaired and didn't get even into the details of what needed to be replaced/what would actually be done. Besides the fact that he asked for more than double the price I paid on both (the Supermicro "died" after a successful BIOS update - everything gets warm when powered up, but no video/etc), as someone that is somewhat versed on hardware like me can't help but feel like I was being gouged. So I'm willing to give a shot to fix'em. 

[Sceptre]

Well, one simple thing you can do is track down the noise that you were hearing.  Presuming you don't have a stethoscope, get a piece of 1/4"-3/8" tubing, hold one end to your ear, and move the other around the board until you locate the source of the noise.  It may be a useful clue.

[janoc]

I sent this same board (and a Supermicro) to a repair shop for a cost estimate; dude told me both would cost ~USD500 to get repaired and didn't get even into the details of what needed to be replaced/what would actually be done. Besides the fact that he asked for more than double the price I paid on both (the Supermicro "died" after a successful BIOS update - everything gets warm when powered up, but no video/etc), as someone that is somewhat versed on hardware like me can't help but feel like I was being gouged. So I'm willing to give a shot to fix'em. 

Of course they quoted you that - because it would be a rabbit hole hunt without any guarantee they could actually find the problem. So you have likely got a quote for a few hours of tech's time poring over the board.

How much is your own time worth? How can you judge that you are being gouged when you don't know what the fault with the board is? The one with the failed BIOS update could be as simple as reflashing the BIOS in an external programmer and it will work again - but it could also be an unrepairable mess if the messed up BIOS caused something on the board to fry itself (BIOS controls quite a bit of voltages on the board too).

[LeonR]

I sent this same board (and a Supermicro) to a repair shop for a cost estimate; dude told me both would cost ~USD500 to get repaired and didn't get even into the details of what needed to be replaced/what would actually be done. Besides the fact that he asked for more than double the price I paid on both (the Supermicro "died" after a successful BIOS update - everything gets warm when powered up, but no video/etc), as someone that is somewhat versed on hardware like me can't help but feel like I was being gouged. So I'm willing to give a shot to fix'em. 

Of course they quoted you that - because it would be a rabbit hole hunt without any guarantee they could actually find the problem. So you have likely got a quote for a few hours of tech's time poring over the board.

How much is your own time worth? How can you judge that you are being gouged when you don't know what the fault with the board is? The one with the failed BIOS update could be as simple as reflashing the BIOS in an external programmer and it will work again - but it could also be an unrepairable mess if the messed up BIOS caused something on the board to fry itself (BIOS controls quite a bit of voltages on the board too).

The fact that they had diagnosed in less than a day both boards and told me that "some parts should be changed" on a the motherboard with the BIOS problem, mentioning nothing about the BIOS itself; and "some parts + BIOS changed" on the Tyan, that didn't even had a BIOS problem to begin with. It's like when you take your car to a mechanic to check a problem, he says you need to do a lot of stuff, costing quite a bit and go to another for a second opinion. I know it is something subjective (consumer intuition), but that's how I take these things on. Given how scarce are professionals that deal with this kind of stuff around here (I need to ship the boards to another state), I decided to give it a go myself.

[LeonR]

So, I've tested the board with another PSU. I found out that If I let the board "rest" for two days or so, it works OK but when I try powering it up the following day the same symptoms occur.

Any input on this would be appreciated.

[janoc]

So, I've tested the board with another PSU. I found out that If I let the board "rest" for two days or so, it works OK but when I try powering it up the following day the same symptoms occur.

Any input on this would be appreciated.

On board power supply/regulator fault? If there is still some remaining charge in the capacitors it won't operate correctly, if you let it "sit" the charge drains away and it works again. That (and the BIOS battery) are the only things on board that could have any "time related" effects.

Of course, that assumes that there aren't other issues and the above is not a red herring - such as your power supply/board being warmed up/cold between the tests, the board being moved (intermittent contact because of a broken solder joint ...) and similar that could manifest themselves in a similar way.

If you rule the above out then the only meaningful thing you could do at this point would be to try to find schematic for the board (likely only on some fishy Chinese forum) and actually start measuring the regulators on the board. However, without knowing what the voltages need to be (schematics ...) and having an oscilloscope to check what the regulators are doing (ripple voltage, etc.) you won't get far.

Did you get that ESR meter and measure those capacitors in the meantime?

[LeonR]

So, I've tested the board with another PSU. I found out that If I let the board "rest" for two days or so, it works OK but when I try powering it up the following day the same symptoms occur.

Any input on this would be appreciated.

On board power supply/regulator fault? If there is still some remaining charge in the capacitors it won't operate correctly, if you let it "sit" the charge drains away and it works again. That (and the BIOS battery) are the only things on board that could have any "time related" effects.

Of course, that assumes that there aren't other issues and the above is not a red herring - such as your power supply/board being warmed up/cold between the tests, the board being moved (intermittent contact because of a broken solder joint ...) and similar that could manifest themselves in a similar way.

If you rule the above out then the only meaningful thing you could do at this point would be to try to find schematic for the board (likely only on some fishy Chinese forum) and actually start measuring the regulators on the board. However, without knowing what the voltages need to be (schematics ...) and having an oscilloscope to check what the regulators are doing (ripple voltage, etc.) you won't get far.

Did you get that ESR meter and measure those capacitors in the meantime?

Not yet. I need to import one of those and it may take from 1 1/2 to 3 months. Postal service here is terribly slow and inneficient. I have about 10 packages that are over a month to be delivered.

I assume the VRMs can be tested as any other component, just need to know which values I'd expect from the testing?

[janoc]

I assume the VRMs can be tested as any other component, just need to know which values I'd expect from the testing?

"VRM" is just a fancy name for a voltage regulator/converter. It is not a "component" per se but a module ("VRM" = "voltage regulator module") containing one or several chips, perhaps some MOSFETs (for the high current ones), some inductors and a boatload of capacitors and resistors. Furthermore, it is very often a misnomer because this circuitry is commonly integrated directly onto the main PCB and not really any kind of separate/replaceable module. So don't expect that you could stick this into some cheap tester that tells you "OK"/"NOT OK" and done.

If you want to test whether it works as it should, you need to know at least what the input and output voltage(s) are supposed to be and what ripple is acceptable. And if those voltages aren't right for whatever reason, you will need to understand how the circuit works in order to have any chance to find the problem. So better start reading up on DC-DC converters ...

[LeonR]

I assume the VRMs can be tested as any other component, just need to know which values I'd expect from the testing?

"VRM" is just a fancy name for a voltage regulator/converter. It is not a "component" per se but a module ("VRM" = "voltage regulator module") containing one or several chips, perhaps some MOSFETs (for the high current ones), some inductors and a boatload of capacitors and resistors. Furthermore, it is very often a misnomer because this circuitry is commonly integrated directly onto the main PCB and not really any kind of separate/replaceable module. So don't expect that you could stick this into some cheap tester that tells you "OK"/"NOT OK" and done.

If you want to test whether it works as it should, you need to know at least what the input and output voltage(s) are supposed to be and what ripple is acceptable. And if those voltages aren't right for whatever reason, you will need to understand how the circuit works in order to have any chance to find the problem. So better start reading up on DC-DC converters ...

Yeah, I meant the components on the VRM, like the driver/MOSFETs, inductors and caps. Since the board have dual socket, I think I can use the other CPU VRM phases as reference. What should I look for? Voltage differences?

[janoc]

Yeah, I meant the components on the VRM, like the driver/MOSFETs, inductors and caps. Since the board have dual socket, I think I can use the other CPU VRM phases as reference. What should I look for? Voltage differences?

That is pretty much pointless if you are not sure whether both sockets don't have a problem - the issue could be "upstream" from them. Or whether the problem is with the CPU regulators at all - not only CPUs have their local DC-DC converters!

The first thing to check would be measure all power rails for each CPU (there are several) and see whether they are in the correct range and not sagging under load. But that will be difficult to do without an oscilloscope - multimeter will not catch any transients that may be on the power rails.

Once you have that checked you can start comparing the two sides but be aware that the the voltages will vary depending on the CPU load. So there most likely will be differences and not necessarily indicative of any fault.