Help me find the short circuit on this WORKING Nvidia GTX1650

[aaronkatrini]

Hi all,

Recently picked up this GTX1650 at a local flea market for a low price and wanted to use it on my secondary PC. I first tested the card and saw the fan spin for a split second and then nothing. Immediately I thought of a short circuit. Disassembled and found a shorted SMD capacitor, injected voltage into it (1V/4A) and two Mosfets got hot immediately (QN3103 & QN3107). Desoldered these Mosfets and the card turned back to life.

Bought replacements from Aliexpress and replaced them afterwards. Tried the card but still the saw the fan spin for a split second and then no power. The short is still there! Removed the new Mosfets and tried the card again, still working fine. I can't seem to find exactly where the short might be. After troubleshooting a bit, beeping every SMD capacitor that could see, I couldn't find much. Even pulled some components just in case (coils, capacitors, crystal)




To keep it short, my problem is this: When the coils are installed in the card, I get short circuit on both the Source and the Drain of the pads in the PCB where the removed Mosfets were, when the coils are Removed I get a short only in the Source but not on the Drain.

I'm basically a noob when it comes to electronics repair but willing to learn. I have access to essential equipment (Soldering station, Hot Air station, Bench Power Supply, Thermal Camera).
Any help would be highly appreciated. Many thanks in advance!

[mon2]

Remove all power. Meter in resistance mode (not diode). Measure the resistance to ground of each of these removed mosfet PCB pads.

So measure the resistance to ground of the source (1-2-3) pin. Then gate (4) to ground and finally drain (5-6-7-8) to ground. Repeat for each of these 2 mosfets. Post each measurement.

[aaronkatrini]

Remove all power. Meter in resistance mode (not diode). Measure the resistance to ground of each of these removed mosfet PCB pads.

So measure the resistance to ground of the source (1-2-3) pin. Then gate (4) to ground and finally drain (5-6-7-8) to ground. Repeat for each of these 2 mosfets. Post each measurement.

Pads of the Mosfet on the left:
Source: 12.74 kOhm - Gate: 25.7 kOhm - Drain: 17.3 Megaohms and slowly decreasing (open)

Pads of the Mosfet on the right:
Source: 0.1 ohm (short) - Gate: 26.64 kOhm - Drain: 12.74 kOhm

[mon2]

Low voltage rails will have a low resistance reading to ground by design.

Make the same measurements on the mosfets that are north and south of these pads. The 0.1 ohms is likely the gpu power rail and so will be the other readings.

[aaronkatrini]

Low voltage rails will have a low resistance reading to ground by design.

Make the same measurements on the mosfets that are north and south of these pads. The 0.1 ohms is likely the gpu power rail and so will be the other readings.

I highly appreciate your time for helping me

Here are the measurements you asked:

The two mosfets North:

On the left: Source: 10.22 Megaohm and slowly decreasing - Gate: OL (open) - Drain: 5.6 Megaohm and slowly decreasing
On the right: Source: 0.1 Ohm (short) - Gate: 18.8 Megaohm and slowly decreasing - Drain: 10.35 Megaohm and slowly decreasing

The two mosfets South:

On the left: Source: 12.22 K-ohm - Gate: 25 K-Ohm - Drain: 14 Megaohm and slowly decreasing
On the right: Source: 0.1 Ohm (short) - Gate: 23.63 K-Ohm - Drain: 12.22 K-Ohm

[прямо]

The mosfets are arranged as HIGH SIDE and LOW SIDE.

HIGH SIDE mosfet Drain pin is always connected to 12V

and the Source pin is always connected to the Drain pin of the LOW SIDE mosfet and also to one pin of the inductor coil closer to the mosfet (while the other pin of the inductor coil closer to the GPU core goes directly to the GPU core) if you lifted the coils, the GPU core no longer receives VCORE and will not turn on.

LOW SIDE mosfet Source pin is always connected to GND.

To check if a mosfet is dead:
Measure its resistance from Drain to Source and from Gate to Source. If you get 0 ohm on both measurements, it is dead.

[прямо]

Also, DON'T MESS with the crystal oscillator for no reason other than water damage/missing components from the oscillator circuit.

Stupid GPU Repair wiki.... it seems to do more harm than good.

[aaronkatrini]

The mosfets are arranged as HIGH SIDE and LOW SIDE.

HIGH SIDE mosfet Drain pin is always connected to 12V

and the Source pin is always connected to the Drain pin of the LOW SIDE mosfet and also to one pin of the inductor coil closer to the mosfet (while the other pin of the inductor coil closer to the GPU core goes directly to the GPU core) if you lifted the coils, the GPU core no longer receives VCORE and will not turn on.

LOW SIDE mosfet Source pin is always connected to GND.

To check if a mosfet is dead:
Measure its resistance from Drain to Source and from Gate to Source. If you get 0 ohm on both measurements, it is dead.

Thank you for your information
What you are saying makes sense in my measurements.
I measured all Mosfets (the 4x ones that are currently on the board and the pads of the two removed mosfets) and there is no short (Source - Drain, Drain - Gate, Gate - Source)

PS, I didn't want to remove the crystal oscillator, it just fell off when applying heat on the back of the card during removal of the Coils

At this point I'm still not sure where my short is. When connecting the coil pads (using a tweezer), thus simulating the coil is still there, I get short circuit from the Drain on the Right mosfets.
When I connect the coil pads of the second coil from the top (L3), I get short circuit on the Drain of the right Mosfet pad (Q13) (Mosfet was removed as stated previously and like shown in the photo)
When I connect the coil pads of the third coil from the top (L2), I get short circuit on the drain of the South Right Mosfet (Q15).

[harp]

Disassembled and found a shorted SMD capacitor, injected voltage into it (1V/4A) and two Mosfets got hot immediately (QN3103 & QN3107). Desoldered these Mosfets and the card turned back to life.

So the card is working, outputing pictures, without two mosfet?
Have you tested these mosfet out from circuit? Both shorted?
Have you check path from driver ic to gate? With dmm can you measure frequency on gate? Or on driver ic outputs?
Maybe is driver ic fault...

It is not helpfull for troubleshooting if you desolder bounch of crucial components... now, you must first confirm that is everithing back again in working condition.

[aaronkatrini]

Hi!!

So the card is working, outputing pictures, without two mosfet?

- Yes, the card is working fine without two mosfets. It loads drivers correctly and can run 3D software. Tested briefly of course to avoid stressing the working Mosfets.

Have you tested these mosfet out from circuit? Both shorted?

- I have tested them, I wasn't sure if they were shorted, ordered new ones from Aliexpress just in case, but the problem persists.

Have you check path from driver ic to gate? With dmm can you measure frequency on gate? Or on driver ic outputs?
Maybe is driver ic fault...

- How do I check it? Gate is not shorted on any Mosfets (or Mosfet pads). Is this "u9510P" chip the Driver IC? Since the card is working I believe that it might be working...

It is not helpfull for troubleshooting if you desolder bounch of crucial components... now, you must first confirm that is everithing back again in working condition

- I desoldered the Coils to separate card in sections to find where short is. Capacitors just in case one of them was shorted. Crystal fell off as I mentioned. I can resolder these components back in place but would be pointless I believe before finding where the short is.

Thank you

[aaronkatrini]

Also wanted to add this, I don't know how I missed it, measuring the pads of the four Capacitors I've removed ( C77, C80, C78, C79 ), I get very low resistance to ground. I get 0.1-0.2Ohms on the negative pad (right side, facing the GPU die) and I get 0.9-1 ohm on the left pad ( left side, facing the video output connectors). Is this normal? Thanks.

[прямо]

I think you are confused and don't really understand what a short circuit is, and what is actually a very low resistance circuit.

The VCORE has 3 power phases, hence 3 same cascaded HS + LS mosfets and 3 same inductor coils. They should have a similar measurement.
So if one phase is different in measurement compared with the rest, then there is something wrong with it.

[harp]

... I get very low resistance to ground. I get 0.1-0.2Ohms on the negative pad (right side, facing the GPU die) and I get 0.9-1 ohm on the left pad ( left side, facing the video output connectors). Is this normal? Thanks.

So, gpu has 0.8 ohm, and your dmm probes and leads about 0.2 ohm...
In this case, 0.2 ohm has to be concider like a short, and 0.8 ohm is gpu working resistance. Anything below 0.8 ohm is suspicious if is on power or signal rail.
Do not inject anything in 0.8 ohm rail.

You ar fascinating with shorts... but, if the gate signal is missing, or fluctating, or pulled aside, it may act like short while is power on. Therefore, driver ic act as supervisor and shutdown rail.

Generaly, signal path is cheking with dmm in beeping mode while one (referent) probe is on what is in interest and another probe searching arround other side of track, seeking to see only 0.000v drop on display (like shorted leads together).
https://m.youtube.com/watch?v=F_Mi-SwZyxY
Inside path may be some components, like resistor, and if it is, just jump them with reference probe and continue... and confirm they are ok.
What is model of your multimetar?

[aaronkatrini]

I think you are confused and don't really understand what a short circuit is, and what is actually a very low resistance circuit.

The VCORE has 3 power phases, hence 3 same cascaded HS + LS mosfets and 3 same inductor coils. They should have a similar measurement.
So if one phase is different in measurement compared with the rest, then there is something wrong with it.

I understand a bit what you're saying. I'm still learning so please be patient
I believe the problem lies on the two bottom phases at this point, still can't figure what my problem is. Thanks.

[aaronkatrini]

So, gpu has 0.8 ohm, and your dmm probes and leads about 0.2 ohm...
In this case, 0.2 ohm has to be concider like a short, and 0.8 ohm is gpu working resistance. Anything below 0.8 ohm is suspicious if is on power or signal rail.
Do not inject anything in 0.8 ohm rail.

You ar fascinating with shorts... but, if the gate signal is missing, or fluctating, or pulled aside, it may act like short while is power on. Therefore, driver ic act as supervisor and shutdown rail.

Generaly, signal path is cheking with dmm in beeping mode while one (referent) probe is on what is in interest and another probe searching arround other side of track, seeking to see only 0.000v drop on display (like shorted leads together).
https://m.youtube.com/watch?v=F_Mi-SwZyxY
Inside path may be some components, like resistor, and if it is, just jump them with reference probe and continue... and confirm they are ok.
What is model of your multimetar?

I need to look carefully at the video multiple times since it is not english and need to figure out what is saying in the subtitles.

I have access to 3-4 DMM, my best ones are ASITA MD591 (daily driver) and Metrix MX22 (backup). Both of them are not state of the art but is all I have and can afford at the moment. Thanks.

[aaronkatrini]

So, gpu has 0.8 ohm, and your dmm probes and leads about 0.2 ohm...
In this case, 0.2 ohm has to be concider like a short, and 0.8 ohm is gpu working resistance. Anything below 0.8 ohm is suspicious if is on power or signal rail.
Do not inject anything in 0.8 ohm rail.

You ar fascinating with shorts... but, if the gate signal is missing, or fluctating, or pulled aside, it may act like short while is power on. Therefore, driver ic act as supervisor and shutdown rail.

Generaly, signal path is cheking with dmm in beeping mode while one (referent) probe is on what is in interest and another probe searching arround other side of track, seeking to see only 0.000v drop on display (like shorted leads together).
https://m.youtube.com/watch?v=F_Mi-SwZyxY
Inside path may be some components, like resistor, and if it is, just jump them with reference probe and continue... and confirm they are ok.
What is model of your multimetar?

I have two chips that could be the driver?
One is labeled uP9510P on the other side of the board, the other is ON 45491 on the bottom (just right of DVI connector), which one is the Driver IC? Thanks

[harp]

Ask google...

UP9510P Datasheet. Part #: UP9510PQAG. Datasheet: 374Kb/19P. Manufacturer: uPI Group Inc.. Description: 3/2/1-Phase Synchronous-Rectified Buck Controller.

Part #: NCP45491. Download. File Size: 153Kbytes. Page: 10 Pages. Description: 26V, 4-Channel Voltage Bus and 4-Channel High-Side Current Shunt Monitor.

So UP9510P is driver.


Video is about fast searching (with copper brush) of connection on burried track over pcb. I show you to see principes of finding connection between two point, as you ask.

MX22 has Hz measurement capability, so it may be handy to see if output have signal to gate...

[aaronkatrini]

Ask google...

UP9510P Datasheet. Part #: UP9510PQAG. Datasheet: 374Kb/19P. Manufacturer: uPI Group Inc.. Description: 3/2/1-Phase Synchronous-Rectified Buck Controller.

Part #: NCP45491. Download. File Size: 153Kbytes. Page: 10 Pages. Description: 26V, 4-Channel Voltage Bus and 4-Channel High-Side Current Shunt Monitor.

So UP9510P is driver.


Video is about fast searching (with copper brush) of connection on burried track over pcb. I show you to see principes of finding connection between two point, as you ask.

MX22 has Hz measurement capability, so it may be handy to see if output have signal to gate...

Thanks, I will study carefully the datasheet of the UP9510P and the layout of the board and see what pin goes where in the pcb and determine if there is anything that doesn't seem right.