Announcement

Collapse
No announcement yet.

Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

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

    Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

    Hello,

    Purchased two GTX980ti palit jetstream to learn something about GPU repair. Both cards came in with shorts on 12V PEX 6 & 8 pin. These are old GPUs so economically not viable to invest time & energy but for sake of learning I am going with it.
    There's bonus of still somewhat relevant performance and SLI if I get them to work. Have all equipment needed for the job as I do repair stuff as hobby.

    Long story short I traced all the shorts and replaced faulty parts. Don't have schematics for this specific board but to an extent concepts are similar vendor to vendor as core/memory have same needs. It seems to have reference power design with 8x DrMOS, 3 phase doublers and ON PWM controller (5V version).

    1st run, replaced:
    1. Power steering mosfets 7403 (switching 12V_F & 12V_PEX)
    2. All DrMOS 6823A
    3. Fuses 2x10 & 1x15A
    4. Measurement after replacement core 3,5 Ohm when cold, mem 71 Ohm. Power rails 12V_F, 3V3_F, 12V_PEX6, 12V_PEX8, no shorts, > kOhm resistances
    5. Vcore 950mv idle, checked PWM pulses on coils with the scope before and after boot to windows, extra phases were turning/on off

    Card booted, windows installed drivers no error 43, all good. Installed GPU-Z all voltages on all connectors normal. Started a game, as soon as card switched to 3d it gave white smoke. PC turned off and the card burnt the same parts.

    2nd run, replaced:
    1. all from 1st run
    2. 3x 81162 phase doublers
    3. 1x PWM controller 81174 - no clear motive why as PWM pulses proven controller is working but had no better idea
    4. Updated vBIOS to ensure no voltage mods
    5. Installed MSI afterburner - set power target to 80% for safety, removed 0 value resistor for OVP & OCP unlock (to lock it)
    6. Checked thickness of thermal paste as resistance drops with heat, maybe due heat issues


    3rd run, thought perhaps 6823A from China are to be blamed, got SIC789 from Digi-Key and replaced all 8 DrMOSes. EVGA used this in their 5V PWM version of 980TI as well, same doubles, same controller, no reason this should not work. But ended with the exact same result. All was going well before starting the game. Game started and after few frames PC turned off. Short is back on 12V_PEX8 DrMOS shorted again (last one). It takes about 1 second for it to die so no way heat is an issue, core hot spot was at 42C during the issue. DrMOSes cooled with thermal pads and stock cooler attached. In fact I am writing this post with one still running... before starting a game. GPU assisted graphics such as in windows or youtube videos work fine.

    There's nothing else I see wrong there other than possibly GPU core itself. Was/is this common? Most references I found conculuded issue was resolved with replacement of shorted parts. Mine (both of cards) produce repeatable failure pattern.


    Regards,
    M
    Attached Files

    #2
    Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

    Further digging produces no results... downloaded the original bios from the card (on the left) and compared to both versions from techpowerup. If anything bios on the card is more conservative in terms of clocks.

    Power tables are identical with limits

    PCIE: 75W
    PEX6: 108W
    PEX8: 175W

    Total: 275W

    6823A can handle 60A each, there are 2 for PCIE (switched with 6PIN), another 2 for PEX6 and another 4 for PEX8. That should be well within operating conditions.

    Never heard of core shorting only under 3d load but have not a lot of experience with GPUs either. Clock scaling in windows works normally. One thing I noticed however is that PerfCap is always idle...

    INA3221 must be working as otherwise where the card would get its power data from to populate GPUZ?
    Attached Files

    Comment


      #3
      Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

      There must be something more out of order. I would assume that in 3d there is much more power consumed buy the GPU, thats probably the difference. Something is disturbing the correct work of the VRM circuit under high load, so that high side and low side mosfets are triggered in the same time leading to a fatal short. Check if the VCC power supply for the VRM chip is stable - usually they have serial resistors 2,2 ohm on VCC, they can loose their parameters if the chip was shorted before, and go to 10 to even 50 ohms, instead of 2,2. This card is also quite old, so electrolytic VCC filtering caps can have lost lot/all of capacitance (usually happens after ca 10 years and more, dependent on the total on-time and working temerature). You can also check whether the gate driver signals on the mosfets look clean (near to rectangular signals, low oscillations), as well as the GPU main power (NVVDD) ripple voltage. Implement some current limitation on the 12 ext supply line to prevent further loosing parts, maybe a fast fuse would do the job, or one or more (in parallel) 45W 12V car bulb in series, but better would be a strong power supply (10-20 amps) with current limit (may be quite expensive however). Not very probable, but would try to exchange the bios chip (they can be partially broken, even if you can read and program them), and/or reflow/reball the GPU (maybe some control signals between GPU to VRM controller have got disconneted with age - cold solder). Last but not least i can imagine the GPU can be broken the way, that engaginng, if a separate one exists (?), the 3d logic inside produces temporary short and overload the VRM circuit, but this is not very likely, as the VRM circuit should activate then the over current protection and protect itself (provided OCP is implemented in this card).
      Last edited by DynaxSC; 04-05-2022, 04:29 AM.

      Comment


        #4
        Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

        These drmos are sensitive to overvoltage on the 5v rail, something to check!

        Comment


          #5
          Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

          Thanks both for taking time to read & respond as it pushed me in the right direction as it seems.

          I started again from checking VCC circuitry and looking back at everything connected to high side as this is what it is killing. Measured all PWM signals from doublers and to the controller, checked feedback pos/neg as these doublers have also power balancing built in so need to process feedback independently and drive DrMOSes on behalf of the controller.

          Mind I work without schematic so probing was time consuming and some errors could have occurred back tracking how things are connected.

          Found 2 things on top:

          1. VCC goes through a resistor Rvcin between Vdrv & Vcin + cap (forming an RC filter) and supplying clean voltage to the logic. Most resistors checked at 10 Ohm (not 2,2 Ohm as suggested but looking at the datasheet 10 is exactly the value mentioned). 2 out of 8 had deviated to 11,7 Ohm and 19 Ohm. Replaced with 10 Ohm 0402 1% and proudly I got really good with soldering 0402 parts

          2. Probably also important as I did not notice this before. Last DrMOS evaporated not only itself but also a capacitor. Tracing it back not any capacitor but what appeared to be its bootstrap cap in series with 0 Ohm resistor between Boot & Phase. That appears essential for timing/driving high side gate which it keeps blowing.

          Circuitry looks identical chip by chip so I desoldered a cap from other phase and measured, 109nF given meter, cables and typical 10% assumed 100nF per datasheet would do.

          Put new parts in and so far it is very promising, played call of duty 4 for 30 mins and nothing burned but power section got very hot. Power consumption oscillates at 240W, load 99%. Now will burn it & test for a while and will check bios with higher clocks.

          Comment


            #6
            Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

            Replace caps

            Comment


              #7
              Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

              Assuming 6 caps on 12V lines? Will ESR test them before replacing out of curiosity.

              Low ESR 105C, will it do?
              https://industrial.panasonic.com/ww/...els/16SEPC270M


              MLCCs and POSCAPs on vcore if not shorted should be good but replacing them all would anyway be a mission impossible anyway.

              Comment


                #8
                Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

                They should od the job for this card, for a newer card I would go for 125C. Pay attention to the dimensions, so that they fit into the space under the cooler.

                Comment


                  #9
                  Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

                  Hey, I'm working on that card as well at the moment, I could help you with a few readings if needed.

                  A side note: how the heck did you get those DrMOS desoldered, are you preheating the board? My 480C hot air station (Yihua) just won't do it.

                  Comment


                    #10
                    Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

                    It depends if yours is still soldered or already welded due shortage, then it will not get off without careful dremeling it out pad by pad. I did not record removal being a problem, alignment of replacement is a bit trickier however.

                    Soldered ones I preheat the board to 90-100C and typically don't need to go to such high temperatures on air station. With pb free solder I do 375-400C depending how it goes with air speed at 5,5 , pb solder 350C usually is fine. My station is 600W 857DW+ so that may be a factor.

                    Comment


                      #11
                      Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

                      I mean you got the possibility of 12v going to 2.5v caps. However everything is working so far so I'd just leave it.
                      Last edited by princeatom; 04-07-2022, 01:53 PM.

                      Comment


                        #12
                        Re: Palit 980TI JetStream 6GB - keeps burning VRM(s) in 3d mode

                        Then GPU chip would be likely dead way before capacitors. Anyway have ordered 16V ones for 12V lines.

                        These 2.5V are 6mm or 8mm diameter? Height 8mm to 10mm will do without issues with the cooler. I mean no problem to measure but not sure how this dimension/footprint is calculated pad to pad, incl. plastic spacer or just aluminium cylinder part.

                        Comment

                        Working...
                        X