Please help with AMD RX Vega 64 shorted

[itzazkrit]

Hi.  I have a reference AMD RV Vega 64 card.   It has served me well mining for many months.  a few months ago, I went to start up system, and no boot.  Now it continues to just turn off the PC like it is shorted.

Things I have done.....

Visually checked PCB for damage power components, etc. but found nothing, no smells, etc.  I'm pretty sure I have identified an issue with one or all of the low-side mosfets powering the GPU.  Please see attached picture.  The IRF6894 mosfets ALL have continuity with ground.  The other VRMs all have some resistance, so I would guess its relegated to one of the 12, or all 12 powering the GPU.

I've powered it up multiple times, and have never "let the smoke out", so maybe it is salvageable?

In its current state, I have applied power, and when I reset the computer, I can get the board to give voltage on the low-side mosfets of about .2V-.8V for a few seconds before, I assume the controller shuts it down because something isn't right.

My questions......

1) Is is safe to assume that one shorted mosfet will cause all of them to be grounded?  Or would it be all that grounded

2)  Likelyhood that core chip is toasted?

3)  Anything else that would cause fets to be grounded across the entire board?

Any help or suggestions on next steps would be greatly appreciated!  Cheers!

[itzazkrit]

I dont know why my orange dots didn't line up correctly.  but they are supposed to be over the physically larger mosfets adjacent to the inductors

[itzazkrit]

high side inductors measure 11.xx volts btw

[itzazkrit]

correction to my original posting.  it seems like they are not shorted but maybe something still is.  I am measuring .89 volts on most of the low side mosfets and .31 on a few.

[extide]

What do you measure, in ohms, between the input of those mosfets and the output? How about from the output to ground?  (Voltage measurements here are not what you want -- you want resistance)

So, yes if one of the mosfets has failed and shorted to ground they will all measure 0 ohms to ground as they are all in parallel. You would have to remove them one at a time and then measure to see when it's not shorted to ground anymore.

Interesting for one of those to fail, though, AMD does use pretty high end parts on the VRM here.

[itzazkrit]

hi.  yes I measured resistance.  the voltage measurements were just after the fact to see if I could find something not correct.  resistance measurement from the case of the low-side mosfets and ground are 0.3 ohms.

voltage measurements were done to see if phases were putting out correct voltages.

[Rerouter]

https://www.techpowerup.com/reviews/AMD/Radeon_RX_Vega_64/7.html

0.9V on left looks ok, would start by looking at the 0.33V mosfets,
They should be IR6894 or IR6211 with public pinouts.

The controller for all the power stages looks to be on the back next to the "place barcode here" text,
on the back of the board near all the power stages is the PWM doublers,

I suspect the 2 big inductors near the 8 pin plugs tie into a common power plane, there should be current sense resistors that you may be able to lift to isolate parts of the board, but I'm not seeing them,

[itzazkrit]

not a whole lot to look on mosfets.  I pulled data sheet for the low side.  visually the ones reading low look exactly the same as the others.  and pinout wise, the case is the drain, so can only measure voltage from drain to ground.  resistance is same between .3V and .9V mosfets

[itzazkrit]

https://www.ebay.com/itm/293047720692

just selling on ebay.  dont have time to mess with.  bid if you want a project.  no reserve!

[itzazkrit]

Thanks for everyones help!

[Rerouter]

https://www.infineon.com/dgdl/ir3598.pdf?fileId=5546d462533600a4015355cdad66176d

The signals to those mosfets go through phase doublers, so yes you can access the gate. and the source at that chip, you may just need some fine wire to probe the pads.

[LateLesley]

Is it just me, or did C116 on the back side of the board look popped??? down near the PCIE x16 connector.

[itzazkrit]

where are you seeing?

[Rerouter]

black capacitor at the bottom left of that image, from the initial rear photo it looks like its cracked on its surface

[LateLesley]

I circled the area, where I think there might be damaged caps.