Seller Note “short on first 8 pin”
- This is accurate, there is a short on the first 8 pin PCI-E
- There is actually a small component (MDV3605 MOSFET) on the Vmem rail that appears to have melted (see picture below).
- resistances seem mostly ok apart from the short
- Vcore – 0.5Ω
- Vmem -108Ω
- PEX – 74.3Ω
- 5v – 12.8Ω
- 1.8v – 900Ω
- I haven’t got a good feeling about this after seeing that low resistance. So far, I have had several 1070s with bad cores, got to be more than a coincidence!
Investigating the Vcore phases
- Measurements reveal that the phase nearest the Vmem (top of the card) has:
- ESR 100, whereas all others are 112
- Gate resistance of 31.8, all others 46
- This DRMOS https://www.onsemi.com/pdf/datasheet/ncp81382-d.pdf is very likely faulty
- Hopefully, it can be removed cleanly and clear the short
- Removed cleanly with preheater and 450 degrees C hot air
The good news is, the short on 12v has gone. The bad news is, there are multiple other issues!
I didn’t notice this damage at first, but I think it could be related to the 5v rail’s low resistance. The damaged component looks like a melted MOSFET (MDV3605 https://datasheetspdf.com/pdf-file/1107906/MagnaChip/MDV3605/1). Although, the low resistance might also be from the phase doubler (https://datasheet.lcsc.com/lcsc/1811051730_ON-Semiconductor-NCP81162MNR2G_C260427.pdf) for the removed power stage (shown above ). I am hoping the flux residue is from my removal of the shorted DRMOS and not someone’s repair attempt.
The voltage situation seems to be variable in that sometimes there has been no Vcore and Vmem and unstable voltages elsewhere. I am guessing this is due to some semi-shorted component that may now be completely shorted. This seems to have settled to the below state now:
- The 5v rail is definitely shorted but still actually has 5v on it! Quite possibly it is feeding another IC that is shorted, probably that melted MOSFET…
- PEX is there at 1v.
- The 1.8v rail looks OK.
- Vcore is 0.7V+ and the core gets hot within a few seconds.
- Vmem is missing, but that is understandable from the above, melted component.
- The backlight is actually coming on on the monitor, which is hopefully a good sign! 🙂
Identifying the 5v short & Vmem 0v
Maybe the next steps should be
- Try to understand a bit more about the circuit containing the damaged component
- Remove Q3602, it is possibly shorted, but certainly damaged
- Check the Vmem PWM, check enable and Vcc (should be 5v, so it’s possible this PWM is part of the 5v problem). Check V ref
Ok so this PWM is https://www.mouser.com/datasheet/2/308/NCP81278-D-1144917.pdf
- PVcc is 5v (unstable)
- EN is 1.8v
- Vref looked like about 2v?
- Measure PWM gates and work out which MOSEFT connects, likely the burnt one is the high-side.
- On occasion, I have notice the 5v rail voltage is unstable and fluctuates around 3v. It’s possible that the 5v buck regulator is also compromised. Check this.