Dead (shorted) graphics card

[old]

Hi, i'm new here and I need help.

Tried to turn on PC earlier today and nothing happened. Then, after disassembling/reassemblin, found out it's a dead graphics card - 12V (6 pin) power supply is shorted.
Nothing looks damaged/burn (at least to me), few fuses I was able to locate/check are ok, and I have no idea what happened - was it card itself (self destruction), or the PSU killed it?

So the question is - can it be repaired, or (at least) can someone help me try (and find out what's happened)? What to check (and how, in what order)?

Here's the card:

[wraper]

If there is a short on 6 pin power connector, most likely some of MOSFETs in step down voltage converter failed short. However I don't think it's worth to be repaired, especially if you don't have experience and hot air station.

[old]

Well at least I can try to fix it before I throw it away... I'm familiar with replacing parts (soldering/desoldering) and some simple electronics, but nothing like this so I really need help finding (narrowing down) "usual suspects"

[wraper]

MOSFETs (circled green), also there is a small chance some 16V polymer capacitor failed (blue). If MOSFET failed, quite likely that PWM controller which drives it failed as well. Note that it's almost impossible to remove those MOSFETs with just a soldering iron, especially without preheating the PCB. They have a thermal pad beneath them which is soldered to the PCB.

[old]

Kind of suspected Q701, Q703, Q901 and Q903 but not the others... So how to be sure without removing them from pcb? What's the PWM, those APW7165C (U801, U901) or?

[old]

Interesting...
S - shorted
x - all of them identical, about 10 times less (30something) than other two (first and last, Q601 and Q901) - tested by placing positive lead on 2/3 pin and negative to 6/7 pin
? - APW7165, I don't know how to test them
?? - SM4912, also don't know

[ogden]

You marked two 16V caps as 0Ohms and one 2.5V cap as 0Ohms. Are you sure you marked right capacitors you actually measure as 0Ohm? How do you measure small resistances, using what exactly tool/model?

[old]

You're right, the 2.5V cap is 1.2 (so ~1) Ohms. It's a DMM - MS8233

[ogden]

You're right, the 2.5V cap is 1.2 (so ~1) Ohms. It's a DMM - MS8233

Understood. If it's not 0 Ohms, don't say so. If you can source replacement mosfet(s) you detect as dead - get them and replace. I would try because current resale/ebay price of card is > 50$

[wraper]

You're right, the 2.5V cap is 1.2 (so ~1) Ohms. It's a DMM - MS8233

That resistance probably is fine and has nothing to do with the fault. With around 1-1.5V on such power rail it's only 1-2W power dissipation.

[wraper]

Unless you have a thermal camera or millimeter with high resolution so you can measure milliohms, you won't be able to find defective component without desoldering it.

[duak]

old, do you have a voltmeter with a 200 mV to 1 V range and a DC power supply that can provide 100 to 500 mA maximum?

If so, set the power supply to about 2 to 3 V and connect to the +12V and 0V or common connections.  Look at the voltages on the tabs of the FETs relative to 0V.  The damaged FET(s) will have lower voltages than the others.  If they are all the same, they are either all bad or some other part has failed.  Go around the card and check the voltages on the capacitors.

Cheers,

[old]

I have 200mV on DMM and DC power with 500mA but only 5V, I'll try to find something more adequate...

But I really don't think they're all bad (all 9 of them), could be 1 or 2 and rest are just "playing dead" because they're connected in parallel or so?

Unless you have a thermal camera or millimeter with high resolution so you can measure milliohms, you won't be able to find defective component without desoldering it.

I'm not against desoldering, just want to avoid desoldering half the pcb (if it's possible).

[wraper]

I have 200mV on DMM and DC power with 500mA but only 5V, I'll try to find something more adequate...

Max voltage does not matter because you won't be able to get high voltage on dead short anyway. Most of the voltage will drop on wires, not graphics card itself where voltage drop should be way below 1V. Best would be to push a few amps from lab PSU and watch with thermal camera what's heating up. Measuring voltage drop on components probably will work but not too well because this PCB has heavy power planes where voltage drop will be minimal unless you push very high current.

[duak]

old, you can use a 10 ohm to 50 ohm resistor in series with the power supply to limit the current.  A 10 ohm resistor will get hot so it should be rated for at least 3 watts.   The actual value will be = (V * V)/R where V is the voltage and R is the value of the resistor.

What electronics experience do you have?  I ask so we can give you a more useful answer.

Cheers,

[wraper]

old, you can use a 10 ohm to 50 ohm resistor in series with the power supply to limit the current.  A 10 ohm resistor will get hot so it should be rated for at least 3 watts.   The actual value will be = (V * V)/R where V is the voltage and R is the value of the resistor.

What electronics experience do you have?  I ask so we can give you a more useful answer.

Cheers,

IMO unless you push 3+A current, you won't find anything.