- The cosmetic condition appears good, although the seller probably concluded that the card has a thermal issue (fans spin at 100% following the crash) and they covered the card with thermal pads.
- Vcore – 0.5Ω
- VMem – 149.6Ω
- PEX – 80.1Ω
- 1.8V – 898Ω
- 5V – 16KΩ
- 3.3V – 34.67KΩ
- 12V – 248Ω! (slot) 46KΩ (8-pin EXT)
- The behaviour is:
- The card starts normally (pre-driver load) and displays fine
- Once drivers are loaded, there is a delay of maybe 30 seconds
- The screen goes blank, the card shuts down power and the fans spin at full speed
- This seems like some kind of overcurrent protection, and that suspiciously low 12V PCI resistance could be a clue
Investigating power delivery
Given the nature of the issues observed, I am wondering if one of the phases misbehaves when the driver’s load. The card is stable initially, so possibly one of the secondary vcore phases has an issue. The primary vcore phase is (shown in blue), and the VRAM phase is of course also active during standby running (shown in red)
I would like to observe the phases using a thermal camera, but I don’t have a reliable way to keep the core temperature down without the cooler. What can be seen viewing the PCB from behind using the thermal camera is:
- Initially, the primary vcore phase can be seen as the warmest spot ~30°C
- When the drivers load, the area behind the VRM phases can be seen to visibly warm up (not drastically ~38°C), whilst the vcore phases all cool down.
Also, the 12V (that has a suspiciously low resistance) powers the dual MOSFET on the VRAMphase, all teh vcore phases use the 12V 8-pin supply.
The dual MOSFET is 4C85N https://xdevs.com/doc/xDevs.com/ocg_1080/NTMFD4C85N-D.PDF
The PWM is Richtek 7J= or https://www.richtek.com/assets/product_file/RT8816A/DS8816A-07.pdf
- Check for anything else on the 12V PCI rail
- Consider replacing the VMem / VRAM dual MOSFET (measure resistance when it is removed, look for donor card)
The good news is that I have hopefully got a reasonable donor card Lenovo GTX 1070 8Gb
I removed the vmem double MOSFET, but the 12V PCI slot resistance is still low at ~135Ω. This means that MOSFET could be fine and the fact that it showed as warm (but not really hot) when vcore shutdown means the problem could be somewhere else.
So, another thing on that 12V PCI rail, apart from capacitors (which if short might more likely make it a dead short), is the 5V buck (U2) converter marked ADP this is actually MPS mp1475dj https://www.farnell.com/datasheets/2915024.pdf
This takes 12V from the PCI on pin 2. One theory could be that the 5V regulator is partially faulty, and perhaps under load (all phases running), it fails and then vcore, vmem and pex shut down? It might be worth removing it and seeing if the resistance improves…
Hmm… well, after removing the 5V regulator the resistance is still only ~140Ω. The search widens.. perhaps a capacitor that sits between two rails could be short?