GPU have all the right voltages but won't boot

[Debidoobs]

Hi Guys,

I just replaced 2 blown fuses on the 12V rail of a RX580 sapphire 8go even though i didn't find any sign of shorts on the 5V, 3.3V or 12V with my multimeter. I don't know the history of the card..

i followed step by step the video from TECH Cemetery () but when on his RX580 sapphire 8go he doesn't have VCORE mine does but won't turn on .. I watched many video but i don't find same situation i m facing. So im clearly lost because all measurements steps i did on the GPU are fine. I have a B550m, a 550W psu, 16go ram...

If i connect the RX580 to the motherboard without the PCIE 8pin connected, the PC will turn on. But fans from the GPU won't start and obviously no pictures on the screen. If i connect the external PCIE 8, nothing will happen, the PSU won't boot.

To be able to make measurement and found i have Vcore, i tested the GPU alone so i use a PCIE riser connected to the PSU. To supply 12V to the 8 pins connector PCIE from GPU, i use a Laboratory PSU. When i start the whole thing, i can see the PSU Lab is delivering 0.4A to the board from 8 pin PCIE. Fans doesn't spin.

Here my measurements (all are exactly similar to the ones mentionned by the video) :

Tab of 5V regulator : 1.6Kohm and 5V ok
1.8V caps : 3.9Kohm and 1.8V ok
DIsplay rail : 32 Ohm and 0.9V ok
Memory controller : 32 Ohm 0.98V ok
Memory : 24 Ohm 1.5V ok
NCP81022 VCC is 4.9V, Enable is 3.3V
Vcore is 0.9V

At this moment, i feel frustration. Note that the ATX PSU will not want to power ON when the GPU is fully connected to it.. We could suppose its because of a short but if i connect a LAB PSU to the external 8 pins, the amp are taken from the 12V !....

If you guys could give me advices it would be so helpful... If the GPU was fried i wouldn't have these measurement so there is something wrong... and also because 12V fuses on external 8Pins where blown in the first place...

Cheers !

 

[1337Goat]

Use a magnifying glass and just trace each trace on the PCB, and make sure everything that should be connected is connected.
Just take the time to check everything, and take notes of all modified chips, or anything that looks out of place.

 

[magda]

Hi Guys,

I just replaced 2 blown fuses on the 12V rail of a RX580 sapphire 8go even though i didn't find any sign of shorts on the 5V, 3.3V or 12V with my multimeter. I don't know the history of the card..

i followed step by step the video from TECH Cemetery () but when on his RX580 sapphire 8go he doesn't have VCORE mine does but won't turn on .. I watched many video but i don't find same situation i m facing. So im clearly lost because all measurements steps i did on the GPU are fine. I have a B550m, a 550W psu, 16go ram...

If i connect the RX580 to the motherboard without the PCIE 8pin connected, the PC will turn on. But fans from the GPU won't start and obviously no pictures on the screen. If i connect the external PCIE 8, nothing will happen, the PSU won't boot.

To be able to make measurement and found i have Vcore, i tested the GPU alone so i use a PCIE riser connected to the PSU. To supply 12V to the 8 pins connector PCIE from GPU, i use a Laboratory PSU. When i start the whole thing, i can see the PSU Lab is delivering 0.4A to the board from 8 pin PCIE. Fans doesn't spin.

Here my measurements (all are exactly similar to the ones mentionned by the video) :

Tab of 5V regulator : 1.6Kohm and 5V ok
1.8V caps : 3.9Kohm and 1.8V ok
DIsplay rail : 32 Ohm and 0.9V ok
Memory controller : 32 Ohm 0.98V ok
Memory : 24 Ohm 1.5V ok
NCP81022 VCC is 4.9V, Enable is 3.3V
Vcore is 0.9V

At this moment, i feel frustration. Note that the ATX PSU will not want to power ON when the GPU is fully connected to it.. We could suppose its because of a short but if i connect a LAB PSU to the external 8 pins, the amp are taken from the 12V !....

If you guys could give me advices it would be so helpful... If the GPU was fried i wouldn't have these measurement so there is something wrong... and also because 12V fuses on external 8Pins where blown in the first place...

Cheers !

Wish you luck OP, I know the feeling...

Is there a place you can send your vid card to have it fixed by a pro? Not talking about RMA

 

[RazorWind]

You have a short to ground on one of the 6 or 8 pin power rails. Check the resistance from the negative side of the fuses you replaced to ground in ohms mode, NOT SHORT DETECTION mode. What I suspect you'll find is that on one of them, you have something like 600-1200 milliohms. If that's the case, you almost certainly have a failed high side FET in the core or memory VRM. You'd then use current injection to figure out which one it is.

If you have something other than 600-1200 milliohms, report back and we'll go from there.

Note that there's a solid chance that you'll need to replace those fuses again, regardless of where the short is.

 

[Debidoobs]

You have a short to ground on one of the 6 or 8 pin power rails. Check the resistance from the negative side of the fuses you replaced to ground in ohms mode, NOT SHORT DETECTION mode. What I suspect you'll find is that on one of them, you have something like 600-1200 milliohms. If that's the case, you almost certainly have a failed high side FET in the core or memory VRM. You'd then use current injection to figure out which one it is.

If you have something other than 600-1200 milliohms, report back and we'll go from there.

Note that there's a solid chance that you'll need to replace those fuses again, regardless of where the short is.

Hi,

unfortunatly, i do have 3.3Kohm to ground from the external 12V power rail ... What i cannot explain is why the atx PSU will not even turn on when all cables are connected to the RX580. The ATX will only turn on if i connect the GPU to a PCIE RISER and feed the external 12V with an "external" 12V supply.. Current is 400mA then, so it behaves normal. Gpu is getting a bit hot but not burning hot.

Im going to check the 12V at the high side mosfet as long as im reading contents about people trying to repair the same GPU. At each 4 VCORE test points (inductors close to mosfet) i have 0.9V, that would mean the mosfet are doing their job.

Also i had a look with a microscope and haven't found any sign of burnt flux under chips as NCP81022 and no shorts over caps around.

If you have some test points in mind

thanks !

 

[RazorWind]

What resistance do you have on the 12V pin of the PCI-E slot connector?

Also, check each of the major logic rails for a short and for function. These cards have at least three. I forget the official names, but more generally, they are:

VCore - Core power. Should be something like 1200 milliohms
VMem - Memory IC power. Should be something like 60 ohms
VMemAux - Memory controller power. Should be around 30 ohms, as I recall. This is usually located between the memory ICs and the slot cover.

What resistance and voltage do you have from each of those.

Also, especially in this context, this device is a graphics card, not a GPU. The GPU is the logic circuitry - a big fat silicon die in the middle, some memory ICs, and some sundry other things. You can generally repair a graphics card. You can almost never repair a GPU.

 

[Debidoobs]

What resistance do you have on the 12V pin of the PCI-E slot connector?

Also, check each of the major logic rails for a short and for function. These cards have at least three. I forget the official names, but more generally, they are:

VCore - Core power. Should be something like 1200 milliohms
VMem - Memory IC power. Should be something like 60 ohms
VMemAux - Memory controller power. Should be around 30 ohms, as I recall. This is usually located between the memory ICs and the slot cover.

What resistance and voltage do you have from each of those.

Also, especially in this context, this device is a graphics card, not a GPU. The GPU is the logic circuitry - a big fat silicon die in the middle, some memory ICs, and some sundry other things. You can generally repair a graphics card. You can almost never repair a GPU.

Resistance from 12V pcie rail and ground give me 4.8Kohm
VCORE to ground give me 1100mOhm and 0,9V ok
Memory controller : 32 Ohm 0.98V ok
Memory : 24 Ohm 1.5V ok

 

[RazorWind]

What do you have to ground on the negative side of the fuses you replaced? Also, are those fuses still intact, or did they go open when you tried to power it up again?

 

[Debidoobs]

What do you have to ground on the negative side of the fuses you replaced? Also, are those fuses still intact, or did they go open when you tried to power it up again?

3.3Kohm , since i replaed them there still intact. I made some progress, well i hope so. I look at schematics and i found that im missing a signal at one input of an AND GATE which is supposed to send a high state (0.3V?) to the PHASE controller in order to start the graphic card. But in my case, PERSTB_BUF still low... don't know where it is supposed to come from yet.
Found this also : https://vlab.su/viewtopic.php?f=48&t=121971
The guy said by changing the U1000 (AND GATE), it did recover the PERSTB_BUF (3.3V)...