MSI RX 580 8Gb Armor

January 14, 2023
 |  2 Comments

Summary

  • I mostly bought this for the aftermarket NZXT Kraken G12 and Corsair H50 liquid cooler. The seller claimed the card is untested. However, I found a low resistance and very hot 5V regulator. So probably just the usual deliberately misleading description. Still, if I can fix the card then bonus time, otherwise no big deal, as the cooler seems to run.
  • Visual inspection appears fine.
  • Resistances
    • Vcore – 1.0Ω
    • VMem – 33.6Ω
    • VDCCI – 21.5Ω
    • Display Rail – 17.4Ω
    • 1.8V – 3.48KΩ
    • 5V – 86.2Ω Very hot APL1117 LDO!
    • 3.3V – 2.4KΩ
    • 12V – 6KΩ+ (slot) 5KΩ+ (8-pin EXT)

Investigating the low 5V resistance

As mentioned, running the card, causes the 5V LDO to get very hot quickly. I didn’t measure the voltages, as the core never gets warm and if 5V is faulty, then no other minor voltage rails should be present.

The datasheet is here http://www.anpec.com.tw/ashx_prod_file.ashx?prod_id=384&file_path=20131021180209364.pdf&original_name=APL1117.pdf

Probably the first thing to do is to remove the LDO and see if the resistance recovers to something more like 508Ω. I actually have an MSI RX 590 Armor that appears similar and has the same LDO, so possibly this can help and maybe be a donor. The ADJ pin is also low at 142.8Ω rather than 384.2Ω.

OK, so removing the LDO hasn’t changed the resistances. I suspected this, as it has been more common in the cases I have seen where the LDO is very hot. The next steps would be to try and identify what is causing the low resistance, there are many possibilities since 5V drives multiple ICs around the card. Voltage injection might help, but it isn’t really shorted at 84Ω, so I’m not sure how effective it will be.

Update 28/02/2023 – 5V resistance increases, but not resolved, the card runs, but has issues…

Well, I tried a few things:

  • Removed the 5V LDO – No change in resistance, the LDO is likely fine.
  • Removed the 1.8V regulator – Again, no improvement in resistance.
  • Injected voltage into the 5V rail – This initially drew a lot of current, like 5-10 Amps. Unfortunately, I couldn’t locate anything hot before whatever it was presumably burnt out! I thought I saw a little smoke, but not much at all. Following this, the 5V rail’s has improved a lot to 482.7Ω (only 25Ω from normal). I still have no idea what was pulling the rail low and cannot see any signs of burns.
  • At this point, I am a bit uncertain of the best next step. Given the resistances are all mostly normal, I decided it might be an idea to try testing the card (the assumption being, perhaps I can spot what failed more easily).
  • I soldered back the 5V LDO and 1.8V regulator – the resistance check seems fine, but still 482.7Ω on the 5V rail.
  • Powering on yields a picture! So I reassemble the cooler and begin more tests.

Testing observations:

  • The BIOS may have an issue (might have got overheated when I removed the 1.8V regulator, took some excess heat I thought), as sometimes the card doesn’t seem to detect properly and was initially recognised as a Microsoft Display Adapter.
  • Drivers can load and the card can run some load. However, the card will eventually blank-screen (caps lock still working).
  • Windows then disabled the card, re-enabling leads to the same issue.
  • When the screen is blank, the voltages appear normal, although VCore remains at ~0.9V (may have seen it fluctuate)
  • One other possible issue under load is that the memory clock in GPU-Z appears to fluctuate e.g. 2000MHz -> 300MHz and back. This could indicate a secondary memory issue? One problem with that Kraken custom liquid cooler is that the VRAM chips get no cooling, which seems bad. I will check the original cooling on the scrap MSI rx 590 I have.
  • tserver gets a certain way, but I think once the card heats up, the blank screens get more common.

Further thoughts:

  • That short was never resolved, the VCore PWM is supplied by 5V VCC, which could be related. It may be getting affected.
  • Memory might be an issue, will try to test this more – passes tserver memfa, perhaps with heat there is an issue?
  • Possibly the core has an issue, although it seems capable of running at full speed at least sometimes.
  • Perhaps there are sporadic power delivery issues related to the 5V rail resistance issues?