Seller Note “Fans spin, blank screen”
Summary
- Some corrosion on the back on the memory PWM
- Resistances
- Vcore – 8.6Ω
- Vmem – 22.6Ω? Samsung, but seems very low, maybe that corrosion around the memory PWM is related?
- PEX – 297 (from U10 GS7103)
- 5V – 10.6KΩ (not sure if this is the right reading, seems high)
- 3.3V – 853Ω
- 12V PCI – 1.55KΩ, PCI-E x 2 6 pin – (J10) 5.2MΩ and (J7) 71.2Ω! (definitely too low)
- There are two BIOS chips and a strange-looking switch
- Oh! There is a bad MOSFET, probably explains the low PCI-E 6 pin resistance
- This card could be a long shot, first to resolve that short, then try to restore the VMem conttroller.
Investigating the low/short 12V rail and blown MOSFET
Hopefully, this will come off cleanly. For reference, it’s gate only measures ~8Ω, but it’s obviously faulty:
For reference, there are 4 phases:
1 x High-Side MOSFET 4C10N – Gate resistance ~35KΩ
2 x Low-side MOSFETs 4C06N – Gate resistance ~29KΩ
There are 4 drivers NCP 81062. And the NCP 81174 PWM
After removal of melted Q12 MOSFET
Thankfully the MOSFET came off quite easily and cleanly with 100 preheat and 400 hot air. The 12V rail has now risen to a healthy 6KΩ++. However, Q22’s gate only measures ~8.5Ω, which indicates further issues, possibly the driver is short too. Since this is very close to the source’s resistance, I measured and found there is continuity between them, obviously not healthy.
After removal of the driver
After removing the driver the gate resistance is now OL as you would expect. Next to try and clean that VMem PWM, I cant really check for a picture until I am confident there is no short there.
Investigating the VMem PWM corrosion
I am not sure how it got like this, but it needs cleaning and possibly replacement. I am pretty sure it’s this one, although I didn’t find that obvious with the 08 marking and no = after it (like with many Richtek ICs) https://www.richtek.com/assets/product_file/RT8809A=RT8809B/DS8809AB-03.pdf
If I can’t clean it well enough, I’ll need to check for a donor card or order one.
Well, my efforts to clean it haven’t exactly been transformative, I only managed to get the worst off.
This IC needs replacing, the pads can then be properly cleaned. Before I invest that time, I will risk a quick power-on test.
Testing and next steps…
Well, maybe there’s hope for this one yet! We have a picture and the drivers load (didn’t test any further on 2/4 phases and a makeshift cooler for the core). At least the picture was clean after driver load, this is enough for me to justify more repair time.
Next steps
- Wait for the 4C10N MOSFETS from china (now have 3 partially fixed cards waiting!)
- Replace the driver (in stock)
- Check all phases with an oscilloscope, there may need to be more replacements
- Remove – clean – buy/get donor VMem RT8809B PWM
- Test further!