ASUS STRIX GTX 970 DirectCU II OC (Card B)

Summary

  • Resistances
    • VCore – 5.6
    • Vmem -79.5 (Hynix H5GQ4H24MFR-R2C)
    • PEX – 292
    • 5V – 3.8K
    • 3.3V – 765
    • 12V – 6K+
  • Fortunately, there is a good schematic for this card
  • Voltages
    • VCore – 1V
    • VMem – 0V
      • VCC – 12V
      • EN – 2.9V (think this is enough)
    • PEX – 0.1V
      • up0132pdda (can’t find a datasheet, but think I get the pins from the schematic)
      • VCC – 5V
      • EN –0V (think this is from VMem, if so, that’s hopefully why it’s shut down)

Investigating VMem PWM

This is the closest datasheet I could find for uP1514P https://pdf1.alldatasheet.com/datasheet-pdf/view/1113860/UPI/UP1514R.html

I now have some spares. Given we have VCC and EN, it seems the PWM could be faulty. I will replace it and test.

Well, the PWM may have had a problem, but has failed on test and now there is a short on the VMem rail. Quite possibly there was a problem with the high-side mosfets. With a short like that, it could be game-over, but I can try replacing both mosfets and the PWM again..

Painful learning, if the PWM is looking faulty, always check or consider replacing the MOSFETS

Well, the replacement seemed to go well and a healthy 11K Ohms is showing on the hig-side mosfet gates. Low gates show 20K Ohms. The sad news is, that I should of knew to check these MOSFETs first of, one was definitely faulty. Now, the Vmem rail is still short, which probably means the overall war is lost, casualties might include core VRAM chips and memory controller. I will try to locate the short by injecting voltage…

1.1V draws 3.3Amps on the Vmem coil. The thermal camera doesn’t seem to show much heat on the core (fingers crossed), but several VRAM chips get warm:

TODO add thermal image

VRAM chip M8 appears the warmest at about 40 deg C, I will remove this first. M3 also seemed warm (several others had no warmth). Damage limitation / possible spare parts assenment at this stage.

OK! So removed M8 ->M3 -> M5, and must have took a bad reading as I thought I saw the short cleared, but it isn’t and doing some more voltage injection shows M6 now at 48 deg C! Maybe this is going to turn out a just of ‘practise benefit’, but it would be nice at least to see if the short is just the VRAM chips.