GTX 1080 TI Memory Corruption \ Replacing VRAM IC

I recently got a non-working Gigabyte 1080ti graphics card for a repair project. The last owner of the card said it was artifacting when it did post and the fans wouldn’t span.


When I got this card it would send a signal to the monitor after a delay in posting but nothing is displayed and having another graphics card in the system I can see the 1080ti in device manager.


After some research I knew what the correct VRM voltages should be, everything tested fine, ran a diode test on the all PCIE lanes going to GPU, found no shorts. The next thing I did is ran MATS/MODS to test the VRAM, I found FBIOD0 bank of memory is the issue.


Replacing the VRAM memory module which I was about to give up to sell it for parts or repair as I have no experience in BGA soldering and don’t have a BGA rework station for it, after seeing many videos on it using a hot air gun I decide to give it a try.

I ordered the correct GDDR5x D9VRL IC for it, this was already re-balled though listed as used and had to get it from a place in China. I was successful on removing the IC using a heat gun then using flux and wick to clean the pads but when I put the replacement IC in its place it appear to be slightly offset from the line markers on the PCB, test the capacitors around the IC, didn’t find any shorts, when testing the card, it didn’t boot like before, MATS/MODS shows as if the IC isn’t connected, I tried a re-flow which got it to shift the IC over which looks correct, MATS/MODS showed write errors though it seems at least it is reading the VRAM, I tried once again just to re-flow but the IC shift too far that I had to remove it which now it will need to be re-balled.

Thinking that some of solder balls were not making good connections with the pads on the PCB or the IC was damage from too much heat or it was bad in the first place.

I was getting errors on D000 D001 D002 D003 D004 D005 D006 D007 for example, does this refer to the data pins from the IC as in DQ0, DQ1 etc?

I was able to fix the fans, the owner before mess with the card had the fans plugged into a RGB header, found the correct header to plug into and now the fans span up.

Sorry for writing a lot though I want to share my experience if it is of any help.

Have any of you try this with just a heat gun and soldering iron re-work station and been successful?

The last removal did do some damage, one of pads pulled off but appears there is no traces to it, might not been used. Might quit on this while it might be still repairable by someone that has better skills and equipment to fix this.

Block Image

Block Image

Beantwoord deze vraag Dit probleem heb ik ook

Is dit een goede vraag?

Score 0

2 opmerkingen:

You mention "The last removal did do some damage, one of pads pulled off but appears there is no traces to it, might not been used" but board is most likely multilayered (sandwiched) so you wouldn't necessarily see the trace.


@aactech you could be correct on that matter however I don't see a via going to that solder pad to lead to other layers, hopefully it is a non connected pad. I tried testing the same pad from BGA area that doesn't have a VRAM IC and didn't get any resistance reading from it to ground, it might not really tell me for sure if it isn't being used. I just wish I was better at micro soldering and BGA re-work.


Voeg een opmerking toe