Broken GTX 980M

Hi,

I am trying to diagnose an error with a broken Nvidia 980 - it shares enough chips with the 980M that I was giving some of the advice in this thread a go.

I'm getting a different reading compared to the first post - when measuring the resistance across the two giant pins on one side of the mxm slot, it starts low and gradually climbs to be hundreds of thousands of ohms within a few seconds. So it doesn't appear to be shorted - but I can't figure out what this might mean?

 

It may be something else entirely other than the MOSFET. Try watching some of the stuffs in this channel: https://www.youtube.com/channel/UCLqBEpeQPZTdhCd0nHWIf6g

If your computer does not POST at all (i.e. computer turns on and immediately turns off again), this could mean your GPU has a short circuit somewhere.
If your computer does turn on, but the card isn't detected, this could mean some components aren't turning on. Think of this as the GPU failing its own POST where something along the chain of voltage rail checks broke.
If your GPU gets error 43, this could be something wrong with the GPU die or RAM module.

For the first case, I would suggest you check the brand and model of the MOSFET chip to see which pins do what. For certain brands like International Rectifier/Infineon, you might need to check the datasheet of a similar footprint MOSFET in the same series instead as I had issue with finding specific datasheet for their stuffs.

 

It is perfectly normal.

To measure resistance you first need to apply a current. For that reason a very small amount of power is drawn from the multimeter's 9V battery and, in your use case, this is now effectively charging the entire MXM board's circuit. Of course, there's not enough power to make the card run continuously but there is enough of it to charge a lot of the capacitors that are on the pcb. Once fully charged their resistance increases and this is what you're measuring.

Conversely, if it didn't show that large resistance value in the end then you'd be looking at a short.

 

What is the symptom of you card, no boot or cannot install driver, etc?

 

I really appreciate the replies. Unfortunately the life of the card may have come to an end.

So the symptoms - it started experiencing a code 43 in windows - i.e. it would boot up, but the 980 was disabling itself and running in a very basic state. I found that if i disabled/re-enabled the card in device manager a few times in a row - it would then stay alive and be detected as a working graphics card! It would then work fine in games and everything 3D related, up until I rebooted. Then it would go back into the basic mode. I guess this was my early warning of problems.

Over time, it eventually took dozens and dozens of attempts of disabling/re-enabling the card in windows. Tedious.

After months of the above, I just couldn't get it to work again and turned to more extreme methods of repair.

"For that reason a very small amount of power is drawn from the multimeter's 9V battery and, in your use case, this is now effectively charging the entire MXM board's circuit. Of course, there's not enough power to make the card run continuously but there is enough of it to charge a lot of the capacitors that are on the pcb. Once fully charged their resistance increases and this is what you're measuring." That would tally with what I was seeing when measuring resistance across the power pins.

So.. even though there was no short - and even though it didn't relate to this post.. I went for broke - pulled the MOSFETs off and tested the resistance of each - all normal. I then reattached the MOSFETs.. and there was a short. I had killed one of the MOSFETs and the card would no longer post.I verified this by pulling off the MOSFETs until I found the one that was now shorted between the two pins that I was originally tested against. My bad.

At this point - I abandoned the idea of repair, and bought a 980M from ebay which I've fitted and got the laptop working again.

Alas I've lost the g-sync with this new 980M.. but I have a working laptop again.

So I am looking wistfully at the 980 sat on my desk, but now see it as a lost cause. I do feel like I have learnt a lot from reading this thread - and even more from the recent replies! Thank you for keeping this post here.

 

Code 43 is most likely GPU Core / Memory Problem, I would suspect dry soldering due to prolonged heat of GPU core, since you dont have shorts at the card initially, it was nothing to do with the power circuit (Mosfet / Caps) the remedial action is likely to be reballing the GPU, the temporary measure is baking the whole card but the problem will be likely to reoccur, anyway good to know you got a replacement and had fun from the old card

 

I have two dead 980M cards , please see pictures.I bought a workstation but very novice at that.
Any reputable place i can send them for fixing?

 

Maybe @Khenglish can help. Depends where you live.

 

Western Canada.

 

The blue looks like some failed SMD caps, might be just a matter of removing those caps and maybe replacing the power MOSFETs.

 

Card 1 should be a pretty easy fix. I have spare parts from dead cards.

For card 2, blown 19V caps sometimes cause internal pcb layers to weld together. This can be sanded or drilled out to make the card work, but it won't look like new.

 

Thanks for the replies , card two is only one month old and the seller is willing to work with me.
I am going to send him both cards to fix.

 

Exact same problem with my alienware AW 17 R1 with a 880m GTX, worked fine but when it died was puzzled for a month trying to fix it took it apart and pulled the GPU out and it booted fine (2016) , was having problems with 8 beep code recently but somehow the unit is operational ???. I cant figure out if the card is shorting the board or the actual slot has some issue. Checked with a multi meter and its 0.3 Ohms.