GTX 980 strix- no image, need some advice in troubleshooting

[andrzejj]

Hello.
Im trying to fix the card that i recently received.
Its asus GTX980 strix
Symptoms:
-No image, screen gets backlit for one second after turning on the PC, than goes into sleep
-It worked in the evening, stopped working in the morning next day, without any kind of magic smoke or magic images.
-After the turn-on, Vram voltage is 1.5V, goes 1.4V in second or two, Vram supply line resistance is 1.5Ohm-seems weird to me, I checked for bad components with IPA and external power supply, nothing found, only gpu gets hot (with voltage connected only to vram )
-Gpu voltage:0.9V, resistance: 1.4V
-Bios chip voltage: 3.4V
-5V line is working and fine
-All fuses tested and fine
-All transistors tested and fine
-Quarz oscillator replaced
After 30 seconds or so all voltages go to 0.
I have no clue what to search for next?
Cheers and thanks in advance for your help

[coldfiremc]

Check for GPU main IC. Is the critical part. In particular, check solder balls. sometimes, cards get twisted because of heat, or improper fan/heatsink installation. Also, mechacnical stress because of partial de soldering is absolutely deadly. Probably you will need some reference

I attached an image of my old videocard. It's clearly twisted and main IC is partially pulled up from the right. Thermal protections were defective and the heatsink was "meh" installed. compare with your card. Take a good photo of it, with a spatial reference(a ruler will suffice) and paste it in some software like autocad. then you can measure how much is twisted. If it's "too much" mm's, probably your card needs some resoldering, but nothing is guaranteed. First think about possible reasons to get the card to that state (bad installation, overheating because of abusive usage...).

If not, check the power stage. Sometimes, those mosfet or thyristor arrays get burnt. You will notice that in that zone, probably the card smells pretty bad there because an old overheating there. Normally this type of failures make protections cutoff "hot" (in the sense of high power) multiphase power stages, so is impossible to detect an overheating, or malfunction, because card is essentially turned off, or in a sort of stand by state. If you detect a "goddamn hot" zone and you have canned dry ice, shot some there and check if this can get card working. If this forced cooling works, try to identify who is getting hot. Measure if you can for ridiculously high currents, in other words, an unbalanced power phase, or just burnt.

Check for cracked power connectors, or solderings on those connectors, loose cables. A spare power supply can help also there.

[andrzejj]

It seems that all balls connect good, around 0,35mm on each side betweend pcb and gpu, i'm worried about Vram resistance- Russians state that 12ohm are normal for zotac's gtx980, I've desoldered Vram power supply mosfets, they seem fine, the resistance havent changed.
And also- buck converter built around AESE chip ( NB671 if anyone would ever search for datasheet) produces 1.07V after the choke (I'm guessing its meant to be 1V, so 1.07 seems OK)
I think i will go full retard and desolder GDDR chips, if the resistance doesnt change-GPU, and if the resistance still doesnt change i will desolder ceramic and polymer caps.
The only thing that got hot when i applied 1,3V to the Vram rail was GPU, maybe internal buffer died short-mode or near short-mode?