Graphics Card Repair - RX 580

[edavi]

Hi all!

I've been into graphics card repair for about a month or so now as a hobby, and I've been picking up a lot of new things about electronics (that wasn't covered in my single EE class they put in our curriculum ).

I've bought about 10 dead cards ranging from old to new across Nvidia and AMD. So far 2 have been repaired, and I'm a little stuck on this next one. I've got a short from 12V to ground on the 6 pin connectors. PCIE seems to be fine. The backstory to this card's previous owner was that a screw fell down the backplate between the PCB. Now I've dealt with a short from 12V to ground on my old GTX 970, and the source was a couple blown MOSFETs - not too bad of a repair, and felt proud of my probing skills.

The card in question is a RX 580 8GB Sapphire Nitro+. I'm learning as I go, but instead of separated mosfets like I'm used to, I believe it's using DrMOS(?) - hope I got that right. So the mosfets are embedded into this power phase? I see no visible damage on them or the card, other than two blown fuses I removed and temporarily replaced with wire. The DrMOS are SIC632. Now please correct me if I went about it wrong, but I checked each for a Vin to GND short internally and saw no issues (same resistance across the board). I've attached images of the card and marked in green with a marker what seemed to have no short (mosfets on the board).

Could I get some recommendations are where to look next in diagnosis?
https://imgur.com/gallery/PDd5GnW

[perieanuo]

short on complex and relatively cheap pcb's cab be succesfuly diagnosed aplying the correct voltage from some variable power supply setting some current limit (like in your case i'll use 12V and minimum 2A), then your 'suspect' heats up
that method is tested in finding shorted capacitors in laptop motherboards, you find the broken one in seconds, can't beat this method (in some laptop mobo I used 10-20 amps to burn the capacitor, seems like a joke but works)
look for some recent radio repair job here on this forum, i proposed this and some joker laught, but finally the OP used the method and found a broken transformer in 2 minutes. the thing, you can't broke what's already broken, and with this aproach you don't make more damage, maybe burn a track bu that's easy to fix afterwards
OR you can take out piece by piece, the visual check often fails short, after 2 hours

[kripton2035]

this tool could also help you to find the short very fast.
https://www.eevblog.com/forum/testgear/finding-short-on-motherboards-with-a-shorty-(with-display)/msg3446526/#msg3446526

[perieanuo]

not bad but i haven't the patience to chech difference with that tool and a milliohmeter
i still prefer my version (everyone already has one good ps in his lab, at least if he wants to deal with pcb's)
but the tool I admit, is less 'invasive' regarding to that poor patient pcb
no offense to anyone, one man can choose his weapon right? 

[edavi]

Update, the exact pin short is on the 8 pin connector marked in red in the attachment. Can this specific pin being shorted point to any hints? Still a short to vcore present. I checked the SIC632 parts and each had no short from Vin to the C-GND, but I'm not sure if I'm checking that correctly. https://www.vishay.com/docs/62992/sic632a.pdf

 I tried applying 12V to the 6pin connector at one of the pins but saw no results heating wise. Am I incorrectly using my power supply and/or should I apply the voltage somewhere else on the card?

[perieanuo]

Update, the exact pin short is on the 8 pin connector marked in red in the attachment. Can this specific pin being shorted point to any hints? Still a short to vcore present. I checked the SIC632 parts and each had no short from Vin to the C-GND, but I'm not sure if I'm checking that correctly. https://www.vishay.com/docs/62992/sic632a.pdf

 I tried applying 12V to the 6pin connector at one of the pins but saw no results heating wise. Am I incorrectly using my power supply and/or should I apply the voltage somewhere else on the card?

can't see red marking, but 8-pin connector is 12V, so almost all pcb takes power from it. you have to trace which step-down IC is 'shorted' or maybe capacitor. if it's the 12V rail shorted to GNDn what happens if you inject 12V from variable DC power supply with for example current protection started from 0 up to 3Amps for example? no component heats up?

[kripton2035]

not bad but i haven't the patience to chech difference with that tool and a milliohmeter
no offense to anyone, one man can choose his weapon right? 

no offence taken, you choose your tools !
but you don't have to check for accuracy of the tool, you only need a visible scale
I don't mind if I read 17mΩ and it is in fact 10 or 25.
I just want to compare to very low ohm values to get the lower one to find the short.

[edavi]

Whoops forgot to attach the image, sorry. Something I've weirdly noticed is that the ceramic capacitors around the top two DrMOS area have lower capacitances than the 4 below. I applied 12V to one of the 12V pins last night, and saw that the top DrMOS isn't getting 12V input. Currently trying to find the source of where this voltage gets stepped down, could it be a faulty resistor?

[perieanuo]

you put wires for fuses, after the fuses you got maybe the zero resistors, check them also.
you may also got protection reverse diodes shorted
all located below nf1206 in your picture ('under' the fuse)

[edavi]

resistors look fine. The R47 parts on the sides of the wire 'fuses' are shorted together. Is this normal? I'm not sure what these parts do.

[perieanuo]

those r47 are just inductances, maybe shock filters after the fuses, should be low resistance

[perieanuo]

if upper drmos is not powered verify also r5174 5169 5170
what do you mean 'tehy look fine'? they measures below 1 ohm on multimeter?

[kavindajd]

You could try powering on the card using a bench power supply, smother it with isoprophyl alcohol and see what IC is soaking up the current as perieanuo mentioned.
You want to make sure you start with a lower voltage <2V to make sure you haven't got 12V running through the GPU core, but maybe on your first attempt the amperage wasn't high enough?

Also what is the resistance of the core?

[edavi]

You could try powering on the card using a bench power supply, smother it with isoprophyl alcohol and see what IC is soaking up the current as perieanuo mentioned.
You want to make sure you start with a lower voltage <2V to make sure you haven't got 12V running through the GPU core, but maybe on your first attempt the amperage wasn't high enough?

Also what is the resistance of the core?

So I've narrowed down a source of heat (burnt my finger) on an IC (1117SA) plugging it into a test bench. The resistance of the core has always fluctuated from around 0.1 to 0.7, but I've heard from sources that the core voltage can be that low? After removing that IC (I think it was for 5V but I can't tell), nothing seems to have changed in the core resistance number.

Also sometimes with my power supply, I set the amps and voltage, and voltage goes through but no amps.. any ideas? Something is keeping the power supply from giving out those amps.

EDIT: Weirdly enough that LDO also made a very high pitched noise when it got 12V. I found a 1117A from an old motherboard and it looks like it has the same deal with ground being connected to its output?

[jmelson]

EDIT: Weirdly enough that LDO also made a very high pitched noise when it got 12V. I found a 1117A from an old motherboard and it looks like it has the same deal with ground being connected to its output?

On the 1117 regulators, the big tab is the output, not ground.

Jon

[edavi]

EDIT: Weirdly enough that LDO also made a very high pitched noise when it got 12V. I found a 1117A from an old motherboard and it looks like it has the same deal with ground being connected to its output?

On the 1117 regulators, the big tab is the output, not ground.

Jon

I measured the resistance and the ground of the graphics card is connected to the Vout pin and tab (also output?), is this normal?

[perieanuo]

EDIT: Weirdly enough that LDO also made a very high pitched noise when it got 12V. I found a 1117A from an old motherboard and it looks like it has the same deal with ground being connected to its output?

On the 1117 regulators, the big tab is the output, not ground.

Jon

I measured the resistance and the ground of the graphics card is connected to the Vout pin and tab (also output?), is this normal?

usually, no
you can measure on 1117 input (batween hic Vin and gnd) and out (batween vout and gnd) and tell us the values
the 1117 heats cause it's broken or his load is broken
of course, you can test the 1117 from your videocard by putting some small load (let say about 20mA) and verify its stabilisation
but be aware, if i remember right, you have fixed and variable 1117 models, cheks yours before, if it's variable he needs some resistor to polarise it, see his datasheet (i guess SA means  it's the adjustable one...)

[debi]

Hello guys,

quite sad this discussion has not been solved, i have the exact same problem about my RX580 that doses not power on, the 12v fuses are open, i didn't find phase problem on the mosfet, the Vcore to ground is  500mohm.. When i try to inject voltage, no amps is taken which is nothing like video repair i can see.. I would like to know if there is a way to check if the problem is coming from the CPU, it would be game over with my tools in this case.

cheers