RX580 No power, is it game over ?

[Debidoobs]

Hello guys,

A friend gave me a GPU Radeon Sapphire Pulse RX580 8gb pretended to be dead. No power at all. I thought it could be a good idea with a bit of research to fix this board regarding to the amount of videos related to GPU repairs saying it's mostly dead mosfet in this case when 12V fuses blow up.. I have a Lab PSU, my multimeter, hot air gun and can solder correctly, also a microscope.. After few readings, i don't know where im going because the problem is nothing like the repairs videos show. In my case, the 2 12V fuses, close to the 8 pins connectors, have blown but i don't find short to ground after this broken fuses...
No short on the 3.3v, aux or 5v rail etheir...

So i tried to inject small voltage on the 12V rail after the 8pin connector but no amp is taken at all....
I found a short between VSS and VDD pins of a NCP81022 chip on the right end side of the board close to the mosfets line.. but not between ground and Vin of the chip.
I tried to find shorts obviously some caps are, some of them on the GPU ic itself but i heard its because of low resistance caps in some area..

I checked the 4 mosfet on the right side, they are no shorted.

I join a picture to my reply with colored dots to give more infos about the board : first of all, I have approx 30 OHM impedance on the yellow dots linked to coils. Fall to 1 ohm on the coils (red dots) close to the mosfets. Then 0,1 ohm on the left of the CMS caps (light blue dots). 22 OHM on the blue dots coil bottom left.

The main thing is i thought i could try to inject low voltage with my psu, right after the blown fuses but he doesn't take any amps, i don't know why.

I would like to know if there is a way to check if the GPU IC itself is dead and it's game over (because i won't be planning a BGA reballing). I heard if you feed voltage on the 12v rail and you have the exact same voltage that goes to the GPU ic, it means it's game over.. but i don't know where to check. If i inject 1V, the Upper mosfet take half of it but the three others take 1V at their Vin.. but once again no amps taken..

Thank you for your help. May be im wasting my time with this board but as we say, live and learn. Electronic is fascinating but it's lifelong road.

Cheers
Geeom

 

[learners permit]

Can't tell by the picture very well but the GPU die appears to be overheated and cracked. Fracture or just not cleaned really well? Clean it with rubbing alcohol and reinspect/retake the pic.

 

[Debidoobs]

Can't tell by the picture very well but the GPU die appears to be overheated and cracked. Fracture or just not cleaned really well? Clean it with rubbing alcohol and reinspect/retake the pic.

Yes sorry for the cleaning, the GPU is fine, no cracks at all.

 

[RazorWind]

Hello guys,

A friend gave me a GPU Radeon Sapphire Pulse RX580 8gb pretended to be dead. No power at all. I thought it could be a good idea with a bit of research to fix this board regarding to the amount of videos related to GPU repairs saying it's mostly dead mosfet in this case when 12V fuses blow up.. I have a Lab PSU, my multimeter, hot air gun and can solder correctly, also a microscope.. After few readings, i don't know where im going because the problem is nothing like the repairs videos show. In my case, the 2 12V fuses, close to the 8 pins connectors, have blown but i don't find short to ground after this broken fuses...
No short on the 3.3v, aux or 5v rail etheir...

So i tried to inject small voltage on the 12V rail after the 8pin connector but no amp is taken at all....
I found a short between VSS and VDD pins of a NCP81022 chip on the right end side of the board close to the mosfets line.. but not between ground and Vin of the chip.
I tried to find shorts obviously some caps are, some of them on the GPU ic itself but i heard its because of low resistance caps in some area..

I checked the 4 mosfet on the right side, they are no shorted.

I join a picture to my reply with colored dots to give more infos about the board : first of all, I have approx 30 OHM impedance on the yellow dots linked to coils. Fall to 1 ohm on the coils (red dots) close to the mosfets. Then 0,1 ohm on the left of the CMS caps (light blue dots). 22 OHM on the blue dots coil bottom left.

The main thing is i thought i could try to inject low voltage with my psu, right after the blown fuses but he doesn't take any amps, i don't know why.

I would like to know if there is a way to check if the GPU IC itself is dead and it's game over (because i won't be planning a BGA reballing). I heard if you feed voltage on the 12v rail and you have the exact same voltage that goes to the GPU ic, it means it's game over.. but i don't know where to check. If i inject 1V, the Upper mosfet take half of it but the three others take 1V at their Vin.. but once again no amps taken..

Thank you for your help. May be im wasting my time with this board but as we say, live and learn. Electronic is fascinating but it's lifelong road.

Cheers
Geeom

What is the resistance in ohms from the bottom side of the fuses to ground?

Note: You cannot use short detection mode to troubleshoot a graphics card. Many of the circuits have a static resistance that is less than the short detect threshold of most inexpensive multimeters.

 

[Debidoobs]

What is the resistance in ohms from the bottom side of the fuses to ground?

Note: You cannot use short detection mode to troubleshoot a graphics card. Many of the circuits have a static resistance that is less than the short detect threshold of most inexpensive multimeters.

Resistance from 12v circuit (after blown fuses) and ground is 3,5KOhm..

Ground to each VRM Coil is 900mOhm.

On the 3 bottom Mosfet, i can find the same voltage in the Vin and in the Phase output.. when i send voltage from my lab psu but no current drawn.
What i can see also is Voltage coming to the Vin are leaving the phase pin from the mosfet so it should be going to the core, if i refer to the datasheet but i don't find this voltage on the coil close to each mosfet...
On the upper mosfet (same column), the voltage at the Vin is approx 300mV when i send 1V. Don't know if it helps.

I also found that the NCP81022 (dual output 4 phase) has his VSS and VDD pins shorted together but not VCC.. don't know if it helps.

The mosfet from VRM core are SIC632. Diode mode on both way between Phase and Pground does not show potential on the multimeter but straight 0 on th multimeter but if i refer to the datasheet, it should be a diode blocking in on way.. i might be pointing something not wrong, because the same thing happen to the 4 mosfet. Still digging and watching video to learn.

I would love to find a way to know if the CPU from the GPU is dead.

UPDATE : finally i sent 0,5V directly at the coil on the left of one of the 4 mosfet and it looks like the Core GPU took 200mA, i haven't stayed long, only 2 seconds. But it looks like the chip is ok if im not mistaken.

 

[RazorWind]

I would love to find a way to know if the CPU from the GPU is dead.

The only practical way to know for sure is to get all of the power circuits running.

One important thing to note: the black square components that you keep referring to as MOSFETs are actually power stage ICs. They have two power FETs in there, plus a phase control IC. This is an important distinction to make because you have two voltage regulators on this board that are made up of discrete FETs and a separate controller.

What the board expects is to have 12V on the Vin pin of the core power stages, and when it's running, you should have around 1V on the switch node (which is connected to the coils, but can only be reliably measured from the left side of them).

When you inject 1V to the negative end of the blown fuses, do you actually have 300mV at all of the Vin pins, or just the very top one?

 

[Debidoobs]

The only practical way to know for sure is to get all of the power circuits running.

One important thing to note: the black square components that you keep referring to as MOSFETs are actually power stage ICs. They have two power FETs in there, plus a phase control IC. This is an important distinction to make because you have two voltage regulators on this board that are made up of discrete FETs and a separate controller.

What the board expects is to have 12V on the Vin pin of the core power stages, and when it's running, you should have around 1V on the switch node (which is connected to the coils, but can only be reliably measured from the left side of them).

When you inject 1V to the negative end of the blown fuses, do you actually have 300mV at all of the Vin pins, or just the very top one?

Thank you a lot to take time to give me informations RazorWind, my bad obvisouly the SIC632 is not a "mosfet".

So to answer you, yes if i send 1V from the blown fuse, only the upper Power stage IC is taking 300mV at the Vin when the three others are taking 1V at their Vin but nothing on the coil..

As previously updated in my last message, i sent 0,5V on the right side of the coil, with current limitation so only 200mA was taken. I don't know yet if i can safely say the GPU is fine. But i ll keep diggin, at least it draws current.

I have to watch the 1.8V ldo ic on the other side of the panel and see his behaviour. I saw it can be the reason why even though the 12V line is ok, there will be no voltage at the coils at all and the car will no power up.

cheers