Help me fix a dead NVidia GTX 780!

[aldfrarug]

Hello there,

just need a few second thoughts for fixing a dead 780. Just want to say in advance I have almost 0 equipment for testing, all I can do is measure resistance, voltage, capacitance and maybe current with an Arduino.

Nothing big to say about the background of the card, I bought it as a defect from EBay. The previous owner report graphics bugs and then the system woudn't start with the card installed. There are quite obvious marks however that this card has been repaired once, the guy I bought it from mentioned he also had it second hand.https://drive.google.com/file/d/16Rlo0ewC-U1XXqa9NZBL8_ciPPhs_7uV/view?usp=sharing

Now let's begin with my results I have gained so far.
At first glance I couldnt see anything defective so I just plugged it in a PC and powered it up. First thing was that two bypass caps behind one of the power stages went smoking. SCP or OCP from the PSU triggered.  Checked them and it turns out they were shorted. https://drive.google.com/file/d/1wkxRdwrz7J94X4WSWdO3Fl6VoL0rqrCH/view?usp=sharing

After I removed them I tried it again and this time an inductor was lit AF. I traced that inductor to the 3.3V supply rail from the PCIe slot. After I removed the inductor and left the pad unpopulated the I measured the resistance between the 3.3V rail and ground aaaaand its 0 Ohms so there is a short on the card between 3.3V and ground. https://drive.google.com/file/d/1loPD1uqZXE1qIMwlLNEnMiVfTWJvoGzN/view?usp=sharing

When injecting 3.3V via the arduino to the card, the system can boot but only with IGP ofc. The system however doesnt even recognize that a PCIe card is in the system. After injecting 3.3V via a voltage divider from the arduino's 5V rail (the normal 3.3V rail can only deliver 40mA) another two caps just below the previous mentioned caps went smoking and the PSU turned off. However this took some time, I was able to boot into Windows with the IGP. I measured them beforehand and they were not shortet. Now they were and I removed them. https://drive.google.com/file/d/16fzsli5KIjWekXCDPp6_a3EmCHumGfy8/view?usp=sharing

After googling all the ICs on the card it looked like only 2 guys are using a 3.3V supply (or at least the ones that I got a result for): the Flash chip and a HS mosfet that is used for the DisplayPort. Thought I'd try my luck and remove the FET, but the short circuit is still there.

I've now installed a 1KOhm resistor where once the inductor was (see img above) so I could see if something was getting hot, but the complete card remains cool. I'd like to note that the fans work normally they neither go 0% or 100% but it looks like the lowest speed they can go.

I also measured the resistance of all the power stages, but none of them seem shorted in any way. When measuring Vcore and VDimm i get ~1V, so there is definetly no dead short between the powerstages and ground.

Whilst doing the pictures for this post I also noted this little guy standing slightly slightly slightly of. Whether or not this was this way from the beginning IDK.https://drive.google.com/file/d/174fu3HLgcgioaQT-Vcsb5ROIN4_Sf6MR/view?usp=sharing

Now I'm beginning to think that maybe the BIOS chip is shorted because the only pin that has a resistance on that chip between VCC and itself is the DIO pin (http://www.farnell.com/datasheets/1725320.pdf that's the chip). On the other hand many pins on there are active low so I might be wrong.

It maybe also possible that a transistor behind the Display outputs is shorted, however i dont think that a single transistor of that size could handle a current that the (former) inductor on the 3.3V rail couldn't.

So it's now come to this that IDK what exactly i should do (I dont really have a second Flash chip lying around so just replacing the chip would take time), so I'm open for suggestions!

Card front: https://drive.google.com/file/d/174fu3HLgcgioaQT-Vcsb5ROIN4_Sf6MR/view?usp=sharing
Card back: https://drive.google.com/file/d/16Z-GzAgv-d6tRbU9tH1xOFpzPjCKmALU/view?usp=sharing

Had to upload the pics to Drive bc the forum doesnt allow more than 5 MB attachments.

In case somebody is interested it's a Zotac Amp!.

PS: Pls don't judge my soldering skills, A 15W iron with bad tips is not really the world.

[Psi]

With zero test equipment available about the only thing you can do is replace every component related to switch-mode power and prey none of the chipset ICs got fried.

But if lots of ceramic caps are going up in smoke is a pretty bad sign. Usually it means one of the switch-mode power supplies shorted 12V into the 3.3V rail and fried everything.

I don't think it's really going to be worth while trying to fix, considering how old a GTX 780 is you can probably get one 2nd hand for pretty cheap if you look around enough.

[aldfrarug]

But if lots of ceramic caps are going up in smoke is a pretty bad sign. Usually it means one of the switch-mode power supplies shorted 12V into the 3.3V rail and fried everything.

I just checked all the powerstages I could find (Vcore and Vmem) for shorts. Found no shorts between the stages and ground were there should be none (There were ofc shorts to the 3.3V line where there was a connection to ground).
The cards using SIC780 powerstages for Vcore and 4901nf for Vmem. Checkd the datasheet for the SIC780s and there shouldnt be any direct connection with the 3.3V rail to the powerstage and Vmem should be running of 12V.

However I noticed another inductor that seems to be out of place (Front pic, the guy labeled1R0 next to the PCIe slot). It's going straight to ground. Is it now a thing to put inductors from a load to ground or is it just connected to the 3.3V rail? IDK. I then checked the IC behind that inductor (back pic, The guy labeled U502) and it turns out that 5 or 6 pins from that guy are connected to ground or 3.3V so I guess that that little fella uses 3.3V bc 6 ground connections seems quite much. So maybe that's the culprit?
Would like to hear your opinion about this bc if I desolder this thing and it is not the problem I would have a great pleasure in putting that guy back there (not).

That IC's labeled 3A FJ GJB, however I couldn't find any information for any combination of that name. I guess it's the IC that powers the Pex rail (a supply for the GPU thats used for PCIe communications or something). Many caps in that area are also showing 0 Ohms resistance to that inductor/to ground/to 3.3V.
If this isn't the bad guy then I would do an oven reflow of the card with my last hopes.

I don't think it's really going to be worth while trying to fix, considering how old a GTX 780 is you can probably get one 2nd hand for pretty cheap if you look around enough.

It's still better than the 560Ti I have rn and I want to wait until the new cards drop until I'm buying a new card so the prices get lower. So I bought a defective card with the hope I could fix it. And the 780 is not that bad, it's about the same as an RX570.