Announcement

Collapse
No announcement yet.

GTX 980 Artifacting even after ram chip swap and core reflow

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    GTX 980 Artifacting even after ram chip swap and core reflow

    Hi Guys

    I have an Asus GTX 980 that has red line artifacts on boot.

    MATS reports write errors on B0 bank (single chip), so swapped out that chip - still errors / artifacts.

    Reflowed the core, same issue.

    Anything else to look at other than it having a dead memory controller?

    Cheers

    #2
    Re: GTX 980 Artifacting even after ram chip swap and core reflow

    no broken traces ni the pcb? voltage normal in that bank?

    Comment


      #3
      Re: GTX 980 Artifacting even after ram chip swap and core reflow

      Originally posted by benwaterson View Post
      Hi Guys

      I have an Asus GTX 980 that has red line artifacts on boot.

      MATS reports write errors on B0 bank (single chip), so swapped out that chip - still errors / artifacts.

      Reflowed the core, same issue.

      Anything else to look at other than it having a dead memory controller?

      Cheers

      reflow means nothing. if u want to really know u must reball. but artifacts is hard error, mostly because of bad trace, capacitor ...

      Comment


        #4
        Re: GTX 980 Artifacting even after ram chip swap and core reflow

        There are principally 3 (4) options: bad memory, bad connection between GPU and memory, bad GPU, OK also a fourth one, missing/decalibrated caps/resistors around the mem chip. If you rule out memory, connections and the surrounding parts issues, then you have the most bad result we do not want to have. So take off GPU and the affected memory chip, check visually with magnifier glas or microscope if tracks/pads under GPU/Mem chip are OK (might be that somebody before already tried to reball GPU and demaged the pads/tracks) and reball/reseat GPU and mem chip. If you have time and the necessary sight and patience, you can check the data/address/control tracks between GPU and mem chip for discontinuity when GPU and mem chip are off board, provided you have a boardview - usual issues are missing pads, track-pad connection, via connections or scratches on tracks. But first assure you exchanged really the right memory chip, they are in different order then in Mats, easy to make a mistake. I would also try to reball/reseat the memory chip again prior to reballing/reseating GPU. Sometimes there are difficult pads (oxidated), and even after replacing memory, you still have the same error. Tin the PCB pads a little to check if they "merge" with the tin correctly, before reballing mem chip and reseating it on PCB. For these operations I'd also recommend to use some aggressive industry flux (eg. KESTER TSF6502), I have the experience that all the easy accessible normal (non-aggressive) non-professional fluxes available in hobby electronics shops often do not make the right job for BGA chips like GPU's. You can also leave a very small "hill" of tin on the PCB pads, this significantly improves the success rate for reballing and reseating big BGA chips on the PCB's (do it not for BGA's with very small and densly placed pads/balls (0,35mm and lower), eg intel chipset's, here the result will be opposite, eg. shorts).
        Last edited by DynaxSC; 07-01-2022, 09:17 AM.

        Comment


          #5
          Re: GTX 980 Artifacting even after ram chip swap and core reflow

          It's been a while and decided to revisit this card.

          After more testing with MATS and flexing the PCB slightly I can change the failing bits within MATS for that bank.

          I am going to remove the core and reball it (and the ram chip) - waiting on a better reballing jig and stencil from aliexpress

          p.s the core was glued to the PCB in all 4 corners - i removed the glue when reflowing it - preheated the board from below and used 45mm x 45mm BGA hot air nozzle from above - got it to temp and tapped the side of the chip and it sprung back into place.

          the ram pads on the pcb and the chip looked fine when i swapped that.
          Last edited by benwaterson; 01-04-2023, 07:55 AM.

          Comment


            #6
            Re: GTX 980 Artifacting even after ram chip swap and core reflow

            While I wait for stencil etc i decided to remove the problematic ram chip again.

            Running mats with it removed results in more artefacts and complete failure on B0 bank (as expected). Pascal cards and above will not display a picture with a faulty chip.

            Whilst the chip is off i thought i would measure the voltage drop using diode mode on the multimeter of all the data lines. Looking for irregular readings or open circuit to diagnose any broken traces or abnormalities in the core.

            So far not seeing anything unusual - 0.575v voltage drop on all lines. However some lines read a little lower unless a press down a little harder with the probe (this is on freshly cleaned and tinned pads) which may indicate slightly oxidised surface.

            Deciding whether to wait and reball the core or reball the ram and reinstall and test first?

            Comment


              #7
              Re: GTX 980 Artifacting even after ram chip swap and core reflow

              Decided to reball and reinstall ram (would have to do it anyway at some point). As it appeared there are no open connections to the core it made sense.

              Still artefacting on test, failing bits now back to how it was originally - 15 to 31.

              Only thing left to try now is core removal and reball etc (not confident this will fix the issue now though - I am just in this to learn more really and see if i can at least remove the core and reinstall back to how it is now).

              Comment


                #8
                Re: GTX 980 Artifacting even after ram chip swap and core reflow

                Ok, got the new stencil and reball jig.

                Reballed and reinstalled core - no change. Still errors on chip B0 (bits 15 to 31)

                Good news is I didn't make it any worse. Reball process took about 45mins - chip came of clean. All good pads on PCB and chip with no oxidisation at all and all traces from that ram chip to the core looked good.

                Just wondering if there is any relationship with the B1 chip that might affect this (that chip hasn't been touched)?

                Comment


                  #9
                  Re: GTX 980 Artifacting even after ram chip swap and core reflow

                  Maybe there is a problem under GPU.

                  Comment


                    #10
                    Re: GTX 980 Artifacting even after ram chip swap and core reflow

                    So there are only 2 options: core is bad or you did not identify the correct ram chip to replace. Are you sure you identified the correct vram chip?

                    Comment


                      #11
                      Re: GTX 980 Artifacting even after ram chip swap and core reflow

                      Maybe a last option: some accompanying devices around the vram chip are decalibrated or so. Eg. resistors with low resistances can loose parameters with time under constant high current conditions.
                      Last edited by DynaxSC; 01-25-2023, 08:33 AM.

                      Comment


                        #12
                        Re: GTX 980 Artifacting even after ram chip swap and core reflow

                        I had to rebuild one pad under the gpu core (not related to this ram chip, was for D0)

                        I am sure its that chip as when testing in mats with it removed (B0, 4th chip anticlockwise from the right) the errors increased and no errors on any other bank.

                        With a bad core i have generally seen errors on at least the full bank (B0 and B1 for example) or all memory banks.

                        I am leaning towards the final option of supporting caps/resistors being faulty - VMEM / VREF voltages did seem to be ok however.

                        Comment


                          #13
                          Re: GTX 980 Artifacting even after ram chip swap and core reflow

                          either that, or the replacement ram chip i installed originally is bad also....

                          Comment


                            #14
                            Re: GTX 980 Artifacting even after ram chip swap and core reflow

                            Originally posted by benwaterson View Post
                            I had to rebuild one pad under the gpu core (not related to this ram chip, was for D0)
                            Ignore that comment - wasn't this card. I am also working on a 1080ti that I have replaced the core on!

                            Comment


                              #15
                              Re: GTX 980 Artifacting even after ram chip swap and core reflow

                              An idea - you could try to use a freezing spray to diagnose the vram chip/gpu, however you need then to mc-guyver some small heat radiator for the GPU, to have access to the vram and not bake the GPU. Artifacts should change, go away or increase with highly negative temperature. So you might finally determine which device is broken.
                              Last edited by DynaxSC; 01-28-2023, 07:49 PM.

                              Comment


                                #16
                                Re: GTX 980 Artifacting even after ram chip swap and core reflow

                                Yeah, I will probably give that a go - good idea.

                                It's odd as only 1/2 the bits on the chip are failing MATS test (all bits fail as expected when chip was removed) which indicates more like a faulty ram chip or connections to the gpu than the gpu itself.

                                Found some good info from northwestrepair (good info in his discord too) - relates to Micron memory, but same principles apply to other GDDR5 / GDDR5X chips

                                Video : https://youtu.be/IBW5alE6K0w?t=345

                                This is currently helping me identify which chips are good from a pile of D9VRL 1080ti chips are have here (found 11 possible good ones from a stack of 21 chips removed from various boards)
                                Last edited by piernov; 01-31-2023, 12:13 PM.

                                Comment


                                  #17
                                  Re: GTX 980 Artifacting even after ram chip swap and core reflow

                                  Ordered some freeze spray - will report back when I receive it and test.

                                  Comment


                                    #18
                                    Re: GTX 980 Artifacting even after ram chip swap and core reflow

                                    Ok, freeze spray made no difference to either ram or core.

                                    I am going to pull the ram chip again and double check connections using diode mode on multimeter (on chip and to the core).

                                    Comment


                                      #19
                                      Re: GTX 980 Artifacting even after ram chip swap and core reflow

                                      Ok, pulled the chip and some very suspect pads (no solder balls on them at all) - very flat dull - see attached image of which pads. Might have found the issue here!

                                      All data lines apart from 2 VSS
                                      Attached Files

                                      Comment


                                        #20
                                        Re: GTX 980 Artifacting even after ram chip swap and core reflow

                                        Originally posted by benwaterson View Post
                                        Ok, got the new stencil and reball jig.

                                        Reballed and reinstalled core - no change. Still errors on chip B0 (bits 15 to 31)

                                        Good news is I didn't make it any worse. Reball process took about 45mins - chip came of clean. All good pads on PCB and chip with no oxidisation at all and all traces from that ram chip to the core looked good.

                                        Just wondering if there is any relationship with the B1 chip that might affect this (that chip hasn't been touched)?
                                        What concerns me is the above - chip has already been reballed to fix this issue - unless both the ram and core needed reballing to fix this and the chances of the only problem on both relating to one ram chip out of 8 chips seems unlikely

                                        Comment

                                        Working...
                                        X