a Novice's attempt at diagnosing faulty batch Sapphire RX 570 4GB itx

9

90Ninety

n00b
Joined
Apr 30, 2024
Messages
4
I am trying to diagnose one of a few identical faulty Sapphire RX 570 ITX card's that I purchased .. I simply thought I could compare measurements ( resistance/ voltages ect ) between a working identical card .

So far for the first card I tried the basic things , such as resistance from 12V inputs (from PSU plug / PCIE pin) , to ground . Then I tried resistance from 12V to all the coils .. These all seem fine and match the working identical card I have

The only difference so far is I noticed that there is a low Ohm's reading (160 Ohm) on the 3.3 V rail to GND . On an identically working card the 3.3 resistance to GND is 470 ohm . Apart form this I have not found any difference between the card's ..

When the faulty card is plugged in , the PC does not recognise it , there is no load going into the card


The only history I know of the card , is that it was previously used in a mining set up

Another card was fixed simply by flashing the Vbios with a lower MemClock ( but it was previously recognised with code 43)


I thought about taking of the VBIOS chip and flashing it .. though not sure

An image of the card is show
IMG_20240502_155254383.jpg


Any thoughts?


Thanks in advance
 
Last edited:
Top right corner plated screw hole i can see green on the plating. This makes me think this has been in extreme humidity and has oxidized off the plating on the copper and you are seeing the copper oxidizing now. Start looking at contacts and pins for signs of damage.
 
Darunion said:
Top right corner plated screw hole i can see green on the plating. This makes me think this has been in extreme humidity and has oxidized off the plating on the copper and you are seeing the copper oxidizing now. Start looking at contacts and pins for signs of damage.
Click to expand...
Well spotted

The cards were from a seaside town near Brighton here in UK, so a lot of humidity .. I could see the heatsink was slightly oxodized .. Though I had seen worse...

This particular card is the best condition card from the batch , and apart from that small spot its in mint condition

I have been over it with a scope a few times , also cleaned it with alcohol


interestingly the only resistance reading that is different is when I checked pin 7 & 8 ( I believe this to be the 3.3V) on the Pcie finger before the notch measures 160 Ohm .. ..On the working card this varies between 450 Ohm and 3K Ohm ( perhaps there is some charge )

Perhaps I should remove the C154 and or C155 , since these are bridged between the pins , I think?
20240502-191846-856.jpg
 
Last edited:
No expert, but 3.3v to ground short sounds like bad news. GPU chip failure maybe. Would have to see what is powered by the 3.3v and start fishing for the blown up thing
 
travm said:
No expert, but 3.3v to ground short sounds like bad news. GPU chip failure maybe. Would have to see what is powered by the 3.3v and start fishing for the blown up thing
Click to expand...
Well I read that also , while its not quite a short , the resistance is lower than the working card

Im not sure what is causing the lower resistance
 
90Ninety said:
Im not sure what is causing the lower resistance
Click to expand...
A path to ground aka a short, that should not be there.

Most passives fail open. Transistors fail somewhere wacky. Most of the transistors are in the GPU.
 
Measuring Vcc to ground, is the resistance constant or increasing over time? Empty caps measure as a low resistance until they build a charge.
 
travm said:
No expert, but 3.3v to ground short sounds like bad news. GPU chip failure maybe. Would have to see what is powered by the 3.3v and start fishing for the blown up thing
Click to expand...
I don't think the 3.3V rail would ever connect to the GPU directly. I could be wrong, but IIRC, on these GCN cards, there are three voltages that actually do connect to the GPU, and they're all 1.5V or less, with regulators on the board for each one.

I'm thinking this is the result of a cracked MLC cap, which usually become a short when they fail in my (finite) experience. Current injection and a thermal camera, or whatever equivalent you have available, is probably what I would try next. Good news is, the 3.3V rail powers only a few specific things on the board, like the BIOS chip and probably some of the VRM controller ICs, so you can probably concentrate most of your inspection efforts in those areas.
 
  • Like
Reactions: travm
like this
You must log in or register to reply here.
Back
Top