Testing motherboard VRM without CPU

[danny1ro]

Hey guys, I'm new here and I'd like to say thanks for all the knowledge on this forum first of all.
I am learning electronics repair, been fixing GPUs and laptops mainly until now, and I'm slowly trying to fix desktop motherboards now.
I had a few successes until now, but I messed things up with one of my recent attempts, namely a Gigabyte H410M S2 with which I burned my test cpu.
I had identified a mosfet which was burned, removed it, cheched for shorts and I didn't find any.
When the new mosfet came, I put it on the board and checked for shorts again and didn't find any. Powered on the board and saw the same mosfet start burning again. The cpu was known good, but when I checked it after taking it out of the board, I found it was now shorted.
Kicked myself for allowing this to happen and started digging around to understand why. I found that one of the drivers controlling the gate of the N channel mosfets that are paired with the burning P channel mosfet now has a short, not dead, but like 2 ohms.

After this long intro, here are my questions:

- When I do finally get that driver I will change it, but is there any way to check if the VRM works without the processor being in the socket? Maybe simulate the processor somehow? I'm trying not to fry another processor if I don't do the right stuff.

- Can someone please explain how exactly that mosfet pair works? I know it's a switching power supply, I know that the mosfets are driven in pulsed mode, but somehow I don't understand this thing fully, otherwise I wouldn't have made that mistake. My mistake basically allowed 12v to go into the processor, but why did that happen exactly?

Thank you!

[DynaxSC]

High side mosfet and low side mosfet must be activated only separately, they are not allowed to be ON in the same time. High side activation is much shorter than the low side activation. This is controlled by the VRM controller, the drivers propagate the signals mostly to pairs of such high side low side mosfet pairs. If your driver had a short it activated probably high side and low side mosfet at the same time for good and this lead to short between 12V and GND, or it activated only high side for good, and CPU has beed fried.

You should have measured resistances of all the Gates if the mosfets before powering the board. If resistances on the Gates are not same and symmetric on all mosfets (VCORE high side mosfets and low side mosfets separately), than something is wrong either with the mosfet or driver or VRM controller (if no drivers present). VCC of the driver can also be measured for resistance, should be not near to 0 ohms. Theoretically also the VRM chip can be faulty, and makes an unintended short circuit signal, but to my experience its seldom.

Dependent on the design many mobos start the CPU VRM without a CPU, but not all. There is a pin in the CPU socket called SOC_OCC (socket occupied), this signal goes to the SIO. If specific mobo does not start without CPU just short this signal to GND at the SIO, so mobo will think there is a CPU.

However this is not the warranty that the CPU VRM will activate power, as the VRMs talk to the CPU to set the voltages desired by the CPU, so if CPU is not present, they might not activate VCORE, or they activate it for a short time, and then switch off again if no communication with CPU possible.

For such tests I buy always a cheapest possible and used CPU, so if it gets fried, the cost is not high. Faulty but not completely shorted CPUs are also good for testing.

[danny1ro]

Yes, not checking the gates of the N channel mosfets was my mistake.
I will test to see whether that pon can help me get at least a quick activation in the VRM, so that I don't risk frying another cpu. It's difficult when you don't know exactly what you're doing.
To this end, l understand that the 1.2v given to the cpu is a direct result of the high side mosfet activating in pulses. This means that some voltage that passes the high side mosfet will be the voltage that reaches the cpu, smoothed out by the capacitor and choke. If what I said is correct, what is the purpose of the N channel mosfets in this game and why do I have only one P channel, but two N channels?

Thank you!

[piernov]

All MOSFETs should be N-channel here, it would be very surprising to find a P-channel in one of these buck converter circuits.
MOSFET driver (or buck controller if it has integrated MOSFET driver) should always be replaced at the same time as MOSFETs, one can cause the death of the other and you never know what failed first anyway, better not take any risk. Also there are many fake MOSFETs on eBay/Aliexpress, better buy from a real electronics supplier whenever possible…
Some controllers may have a pin that can be set in order to turn on the VCore at some default voltage even if there is no communication on SVID, need to read the datasheet for that.

[danny1ro]

You are correct, the mosfet is RA88 and is an N channel one. Why is it an N channel, and why do I have the other two, which are also N channel?

[DynaxSC]

See this:

https://www.personal-view.com/talks/.../21417/motherb

[danny1ro]

That's by far the best resource I came across that explains a VRM. Now I understood what each of the components does. I looked up why N channels are there and cleared that one out as well. Now all I have to do is wait for that IC to arrive and hopefully not blow another CPU
Thanks again for all the help!

[DynaxSC]

Good luck, and write about the outcome.

[anikin73]

Burning mosfet again is a problem that will occur. How do you know is the VRM chip normal ? Is it possible to know is the VRM chip faulty ?
there are 2 issues here.
1) burning the cpu .
2) blowing the mosfet again.

answer to 1. If u do not put cpu , you cannot on the motherboard unless u are using MSI motherboard. Msi motherboard has a bypass pin to do it.
others need bypass SKTOCC signal at IO chip to cheat IO that there is a CPU.. then u can run without cpu.
answer to 2 . How to prevent mosfet burnt again if the problem is with VRM chip. I went through this before. Now I had a motherboard that ram mosfet burnt. I suspect is the vrm chip generating the wrong voltage. So either I buy a few more mosfet to try or I use other signal to try to activate the mosfet and measure the actual voltage from VRM.

[DynaxSC]

Additionally u can remove all mosfets in the burning section (both HiSide+LowSide if there is no doubling, and 2 x HiSide+LowSide if phase doubling is used) and start the board w/o CPU as anikin73 wrote. Then look with a scope what happens on the gates of the mosfets, and/or the phase outputs of the VRM/Driver/Doubler. The board might work even without one phase if CPU voltages are correctly present, also if u insert the CPU (but I strongly do not recommend to do this).

Also a faulty VRM/Driver/Doubler can have have different resistances on the phase outpus/inputs (eg. some internal short), but not necessarilly. Other method is to compare VCC resistance of the VRM chip, a faulty one can have a different resistance on VCC, but u need a second board to compare this, or a new VRM/Driver/Doubler chip.

In this case u should also measure all low ohm resistors (up to 100 ohms) and caps in the VRM section, caps can go shorted, low ohm resistors might burn in certain cases and have a much higher resistance than nominal, or even can disconnect totally. If there is such one faulty device on one phase it might lead to the mosfet short. Low ohm resistors can be usually measured in-circuit w/o big error, higher value than 100 Ohm - 1kOhm may show different values in-circuit.

What VRM model is it ?

If it is a programmable one, no configuration/firmware files are available, special HW and software are necessary to program the VRM chip (I2C interface), and generally only chance to repair is to use a chip from a donor board.

[danny1ro]

Hey guys,

My vrm chip finally cam from china and I soldered it on the board. It's an ISL6625ACRZ.
Measuring resistances on the mosfets I still think I'm getting some weird readings.

Case 1- measuring without processor in socket:
- high side source: 99ohms
- low side drain and source: 99 ohms
The gates measure 2 - 3 kohms.

Case 2 - with processor:
- high side source: 0.7 ohms
- low side drain and source: 0.3 ohms

Somehow this seems a bit low to me, but the weird thing is that I am getting this on all the phases once I put the processor in. This is a known good processor.
I tested the caps inside the socket and the all measure 99ohms.

What do you guys think?

[anikin73]

It is normal. With processor, the low ohm is because it is directly connected to processor, hence very low like shorted. It is your gate measurements. You worry short / kill cpu, you have 2 choice. Get a spoilt cpu to try, Also you might wanna change both upper and lower mosfet together.

[danny1ro]

Thank you for the help guys, the motherboard is fixed and nothing caught on fire .
It was that little chip that controller the mosfet, too bad I had to burn a new mosfet and a cpu to learn this.
The low resistance I was measuring is ok.
On to the next one!

[sam_sam_sam]

Thank you for the help guys, the motherboard is fixed and nothing caught on fire .
It was that little chip that controller the mosfet, too bad I had to burn a new mosfet and a cpu to learn this.
The low resistance I was measuring is ok.
On to the next one!

I bet this was a great learning experience and opportunity but it probably the best education of how to troubleshooting a motherboard and it supporting components
Great job repairing it and now it is working again

[danny1ro]

Thanks, indeed it was. My next one is a B660 which doesn't start... we'll see when I get around to it.