Sapphire RX 580 no Vcore repair

[DeVillager]

Hello,

recently my friend brought me his Sapphire RX 580 graphics card that stopped working and told me to repair it. It's not being detected by PC at all and when installed as a secondary GPU Windows won't detect it either. However, what's weird is that every few hours it just starts working normally, I have ran a stress test for about an hour and used it for a few more before it crashed and refused to work again. This cycle keeps repeating.

I took it apart and while I couldn't find any shorts or blown fuses I did discover that Vcore voltage was only 0.25V. All other voltage supplies were present on the board so this led me to start here and check the VRM controller(NCP81022). There I found that there is no 3.3V on pin 10(enable pin) pin 9 on the VRM controller that can be measured there at the times that GPU works normally.

Ultimately, this is where I got confused and stuck and came here to look for help as I am not sure where exactly does enable pin get 3.3V from.
Also, enable pin pin 9 is connected to some other hard to find component(U1000) which Russian forums say is an AND gate as can be seen on the attached picture. Red spots measure 3.3V and green measures 3.3V when GPU works and 0V while it doesn't. Output of U1000 goes to pin1(PWROK) of the VRM controller.


I would appreciate if anyone could help me find out where 3.3V for pin10 are supposed to come from or if you have any other suggestions to further diagnose the problem.

Thank you!


Datasheets:
NCP81022: https://media.digikey.com/pdf/Data%20Sheets/ON%20Semiconductor%20PDFs/NCP81022.pdf
AND gate: https://www.ti.com/lit/ds/symlink/sn74lvc1g08.pdf?ts=1627282215175&ref_url=https%253A%252F%252Fwww.google.com%252F

[DeVillager]

I made a mistake earlier, the pin I was talking about as 10 is actually pin 9(VDD_POWERGD), pin 10 is getting 3.3V as it should. 

Now, this makes much more sense than before as datasheet says pin 9 is an output(VDD_POWERGD - "Open drain output. High output on this pin indicates that the Main Rail output is regulating.") that is giving signal to U1000 AND gate input pin 1 when GPU works, when not there is only 0.015V there or so. Input pin 2 of U1000 always gets 3.3V so I didn't bother finding where it comes from exactly. So when both U1000 inputs are present, U1000 will output PWROK signal to pin 1 of NCP81022 and tell it everything is in order with the power delivery system(?).

As I have said, all voltages are present in the gpu, no fuses blown and NCP81022 is getting power but is for some reason outputting only a tiny voltage from pin 9 and only giving slowly fluctuating 0.25-0.35V or so to the GPU core while GPU refuses to power up properly.
So for now I blame NCP81022 because I think it is the component that failed. Vcore is being controlled very wrong and everything else I checked on the board looked good.  That or some part of power sensing circutry that is giving wrong information to NCP81022 so it thinks it's not safe to send power to the GPU core.

My knowledge and tools are very limited so if anyone has any ideas don't hesitate to write.
I will update this thread if new discoveries are made or chip is replaced.

[UZI175]

Hello.
Im having the exact same issue with a Sapphire RX 570 4G ITX, did you solve the issue?