Nvidia GTX 780 PCB circuit / component ID

[HamBown81]

Hello, I am new to the forum and could certainly use some assistance in repairing my graphics card.

I found a blown SMD, which I believe to be a multilayer ceramic capacitor, but it is in a 0402 package and I can't find any schematics to aid with identification.  I have included a couple of photos of the circuit.  Any assistance would be greatly appreciated!

[Rasz]

what makes you think its bad? did you desoldered/measured it?
have you diagnosed the card? measured all the relevant rails?
this will give you some idea about going about it https://www.youtube.com/channel/UCRRTfkRDKqsI9FSXP1sc4kw/videos?disable_polymer=1

[Bashstreet]

Yes it looks like it has heated up (from discoloration around the component) and seems to have blown a crack in it)

Question really is what made it blow any case i cannot say i can identify it more than it looks like ceramic capacitor.

I do not think the capacitance will be critical (some extra capacitance should not hurt) but you should get high voltage one as possible.

Doubt even running it without the cap will be a problem and you could rule out the problem that way ... although if something caused the cap to run hot (voltage or current) it can pop the next cap resistor or even worse.

Maybe someone can give further advice.

[HamBown81]

I was unable to get a good photo of it with my phone.  The package of the SMD has significant physical damage and is discolored.

The card will boot Windows and is displaying graphical artifacts that would indicate a vram issue. 
The fan runs at full speed and properties do not display properly in GPU-Z.
The vias just above the component are the back of the GDDR5 module, closest to the VRM.

I will be doing some electrical testing later today but I am fairly certain that this cap is either wholly or partially responsible for my issue.

I am certainly a noob, but I am learning as much as I can. I have some rework and soldering equipment and a bit of experience as well

[HamBown81]

In addition to my previous reply, the card was being run as a hybrid with an AIO water cooler. I suspect there may have been liquid damage in that area but I don't know for sure.

Everything else looks perfect on the card. There was just some blackness on the components beside this one but I assumed that it was from it blowing up. Either that or there was a drop of liquid there and it turned into blackness as it got cooked to death with the remnants of whatever blue smoke the cap let-out when it gave up on life.

[DaJMasta]

Have you looked at reviews of the card?  I know there are a couple hardware sites that will give you detailed board shots for the purposes of voltmods and stuff, could give you a reference to compare to for help figuring it out.

Giving it a measure is probably a good idea, and if you're convinced it's dead, maybe removing it altogether would help get a better look.


Looks like a cap to me in all pics except the last one, where it almost looks like the case is translucent, in which case I'd sort of expect either an LED (but it's likely too small for that) or an inductor... hard to say.

[Rasz]

I was unable to get a good photo of it with my phone.  The package of the SMD has significant physical damage and is discolored.

it looks oxidized, I dont see cracks, are there any? I would start by measuring it (is it shorted?) and  reflowing

The card will boot Windows and is displaying graphical artifacts that would indicate a vram issue. 
The fan runs at full speed and properties do not display properly in GPU-Z.

if I had to guess I would say U502 is monitoring gpu voltage rails, that would explain both lack of sensors in gpuz and full speed fan
that would be logical, but u502 is a pcie supply, afaik 0.9V, rt8071azqw
corroded capacitor is C639

AIO water cooler

all the brown oxidized vias (should be covered with black soldermask) show signs of water damage, including the ram ones. This is 8-10 layer pcb, might not be fixable if vias corroded deep enough.

[HamBown81]

Whatever it is, it is short. I have measured again, and the resistance is 1.1Ω across (with 0.06 being a full short). I will try to get a better picture later-on.  One side looks like it blew-up and is lifted from the board, the component is definitely bad.

Regarding the vias, the color in the photo may be deceiving; I don't believe they are corroded.  The second photo is the closest to real-life, but the colors are still wrong due to the LEDs on my microscope light.  The vias are the same color (of copper) as the mounting holes nearby; and, the whole card looks like that on both sides.  IF there was liquid damage, it was only a drop in the area of the blown SMD.

[HamBown81]

The photo is bad, particularly the last one.  I took it through the eyepiece of my microscope, with my phone...  I really need to get a proper camera (I have a simul-focal AmScope set-up).

It SHOULD look like a smaller version of C636, the package is the same color as well.

The case does look a bit translucent, but it got cooked.  It looks something like this, from the side

[HamBown81]

it looks oxidized, I dont see cracks, are there any? I would start by measuring it (is it shorted?) and  reflowing

The one side looks like it got firebombed, or torpedoed... It is missing pieces and is significantly lifted from the PCB (while still sort-of attached).

u502 is a pcie supply, afaik 0.9V, rt8071azqw
corroded capacitor is C639

This is a representation of the field of view, on the other side of the card.  I would think that PCI-E should be on the opposite side (where the 12v slot power is located?).  My initial thought is that this area was voltage regulation for VRAM or, potentially, PEX/PLL based on location?  Of course, I could be totally off-base...




Here are some basic measurements of resistance, as well.

[HamBown81]

I drew some of the "fences" back on here, so that the traces are easier to see.  The SMD appears to be between the GDDR5 module and GND.




These cards, apparently, are known for blowing-up their memory VRMs and inductors?  https://youtu.be/6BBEasRiM_M

[Rasz]

just desolder this cap, I would be careful not to take it with pads
u502 is pcie controller supply vrm, output is on L5, card wouldnt work at all if it went bad so this cap has nothing to do with it, might be ram chip decoupling

These cards, apparently, are known for blowing-up their memory VRMs and inductors?

L12 L13

[HamBown81]

Thank you for your help.  I will desolder the cap and continue testing.

I did manage to take a much better photo after cleaning the card.  I believe you are correct about the corrosion on the vias.  I am still not sure why it is so ubiquitous across the card.  Could the via corrosion be caused by high humidity, rather than traditional liquid damage?  Please ignore the random bit of red debris in the lower right, it is incidental.