Seller Note “PC doesn’t turn on when the card is inside PCI-E x16 slot”
Summary
- The condition seems OK, was cheap, and appears to have a 12V short on PCI-E 8 pin
- Resistances
- VCore – 4.6Ω
- VMem – 81.8Ω
- PEX – 348.4Ω
- 5V – 506.9Ω
- 3.3V – 914Ω
- 12V – KΩ++ (Slot) 4.6Ω! (8 pin) 55KΩ+ (6 pin)
Investigating the 12V short
It could be a bad sign that the 12V resistance matches the core resistance, but let’s work it through. Seems like decent scrap worst case, I have two other cards like this one.
The below phase U4540 is my immediate suspect, see the solder balls circled in red. Further confirmation of an issue is the high gate resistance (R283, Orange square) has a very low value of about 40Ω, which should be 7.98KΩ.
After a luckily clean, removal, the gate resistor now measures a healthy 8.5KΩ (still a bit too high good to check the high-side driver) and the PCI-E 8 pin filter is now 55KΩ+, but… is the core fried?… good news! There is a BIOS picture at least and the remaining phases appear clean 🙂 I have some spare SM7320.
The DRMOS area needs cleaning up and some fresh solder first:
I have replaced U4540, looks OK, but I am still not happy that the high-side gate measures a bit too high. I have had this sort of problem before, and it could be a sign the high-side driver was the reason this DRMOS shorted before. Simply trying it could burn it out again, possibly damaging the board if I am unlucky.
news flash! I discovered my other multimeter’s probe was not properly screwed in, now all phase’s high-side gates measure at 7.98KΩ. Time to cautiously test…
Testing
Well, the good news is there wasn’t any smoke and all phases appear to be working!
- GPUZ render test appears fine.
- Kombuster HD passes, and the core and hotspot temperatures are pretty good.
- VDDC of 1.218V seems high to me, but it could be normal for this card.
- This card doesn’t run it’s fans much (come on sporadically with spikes in temperature and don’t run that fast under full load), so they are maybe linked to core temperature.
- I am paranoid about whether there is some underlying VRM efficiency issue, it seems very hot to touch the back of the PCB behind the VRM, this is quite possibly normal for this type/age of card.
- I will do a longer test using Subnautica and take a look with the thermal camera
- Core temp of 60°C and hotspot of 75°C seem good
- The back of VRM area on the PCB is hot at ~92°C, again, I am not sure yet if this is bad.
- I need to compare it to the other MSI GTX 970s I have
When running Heaven HD Ultra sustained
- The core is flat-out 99% load, but only ~60°C with 75°C hotspot
- That back of VRM area tops 100°C at times
3D Time Spy scored about average.
Next steps
- Try to get some comparisons – should VDDC be 1.218V? Seems very stable at that level on the graph
- Maybe the VRM heatsink thermal pad needs replacement, I don’t imagine this will change much though
- Continue testing