Announcement

Collapse
No announcement yet.

GTX 780tI Repair attempt...

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

    GTX 780tI Repair attempt...

    OK guys so I rightly or wrongly want to repair graphics cards. To this end I bought three GTX 780ti off ebay as Spares or repair to see what I could learn.

    They are different brands, one Palit Jetstream, one Gainward Phantom and a Gigabyte Windforce

    I selected this model as faulty cards can be had quite cheap (around £20 each) and working ones hold a good resale value being one of the best cards that supports Winpows XP for retro gaming.

    So out of the three I have two that don't detect at all and one that detects but gives no video output

    I'm looking at the Palit first. This one does not detect by the BIOS. There is a problem with what I thought was an inductor L17 marked 'X' (but apparently it is a 12A fuse?) being open circuit.

    I took some resistance measurements across the capacitors an the 12V input and low voltage ouput side of the VRMs , there is evidently a short somewhere.

    The attached pics will describe this better than text I could write. The reistances marked 'high' are because I just see capacitors charging up so it is hard to get a proper reading

    Does someone have experience of these cards to know what I am looking at here?

    Are there any resources (board views, schematics or other info) that would help me?

    Cheers
    Rich
    Attached Files
    Last edited by dicky96; 04-17-2020, 06:30 AM.
    Follow me on YouTube
    ------------------
    Learn Electronics Repair
    https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

    #2
    Re: GTX 780tI Repair attempt...

    I had a quick look at the other two cards with my multimeter. One of them (I think it was the Gainward Phantom) looks to be very similar to the Palit card

    On this one both inductors/fuses? L17 and L18 read OK. I don't have any shorts on the electrolytics on the 12V side of the VRMs, and on the low voltage side I have 3R3 across all the capacitors that read 2R4 on the Palit. The other two capactors at the top of the card on the low voltage side that read 0R on the Palit read about 110R on this card. from this I am pretty sure there are two Voltage rails generated by the VRMs, with more phases on the one that read 2R4. Which makes sense.

    The other one seems to be quite different in its layout and components. On this one I don't see any shorts on the electrolytics on the 12V side of the VRMs but I read near to 0 ohms on all the low voltage side electrolytics.

    Comparing all these facts, possibly the Palit has two separate faults - a short circuit MLCC or Mosfet or similar on the one of the 12V rails, and a short circuit capacitor or load on what may be the Vram supply?

    The Phantom reads no shorts anywhere. This may be the one that detects in the BIOS but does not give any video. It may help me to determine what voltages should be present on the other ones.

    The Gigabyte (I think it was) may have short circuit capacitors or VRMs, or both GPU and RAM short- this just seems to be short everywhere on both the low voltage rails.

    Still if anyone has more experience (and that would not be difficult as I have zero on GPUs) I hope you can advise me.

    Rich
    Last edited by dicky96; 04-17-2020, 09:35 AM.
    Follow me on YouTube
    ------------------
    Learn Electronics Repair
    https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

    Comment


      #3
      Re: GTX 780tI Repair attempt...

      I spent a little bit more time looking at this

      I removed two inductors from the Palit board. I then found the short circuit on the 12V side (L17) had gone away.

      I still have a short on the low voltage side of those inductors. I checked and one end of the coil goes directly to the L17 and the other end to the MosFET drain, which I thought is a bit odd as this therefore is not a buck regulator type circuit as I expected, it's more like a PFC type topology??

      Once I got rid of the short on the 12V side I can see that there are two FETs, Q15 and Q16, driving the two inductors and drain of one of them goes to L17, drain of the other the other one goes to PCI-e 12V

      Also it turns out the GPU on these two boards are different. Maybe that accounts for the 2R2 vs 3R3 resistance I can see on what I assume is the BGA supply rail.

      The Gigabyte card is also GK110-435-B1 same as the Palit. This one reads short on both the low voltage rails. I thought all these cards were supposed to be GTX 780ti - but maybe that is not the case.

      The pics will explain this better

      I'm thinking now to power up the Phantom and get the voltage readings for both the low voltage rails. I then could possibly use my bench PSU to see if I can find the short on (Vram)? on the Palit. But would that damage the GPU if it was not also powered? Assuming the GPU is actually still OK?

      Rich
      Attached Files
      Last edited by dicky96; 04-18-2020, 06:05 AM. Reason: more info
      Follow me on YouTube
      ------------------
      Learn Electronics Repair
      https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

      Comment


        #4
        Re: GTX 780tI Repair attempt...

        GK110-300-B1 GTX 780

        GK110-425-B1 GTX 780 Ti https://www.techpowerup.com/gpu-spec...ia-gk110b.g760
        All donations to badcaps are welcome, click on this link to donate. Thanks to all supporters

        Comment


          #5
          Re: GTX 780tI Repair attempt...

          Originally posted by SMDFlea View Post
          GK110-300-B1 GTX 780

          GK110-425-B1 GTX 780 Ti https://www.techpowerup.com/gpu-spec...ia-gk110b.g760
          OK thanks. The one that reads the lower resistance is the 780ti, which kinda makes sense to me. So maybe both those GPU are good.

          I just realised the MosFET 4901NF is actually a dual MosFET with 10 connections (2 more pads underneath) so the toplogy of the converter circuits around Q15, Q16 may not be what I thought. I will try to reverse engineer that part.

          https://pdf1.alldatasheet.com/datash...LLD4901NF.html

          Rich
          Follow me on YouTube
          ------------------
          Learn Electronics Repair
          https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

          Comment


            #6
            Re: GTX 780tI Repair attempt...

            Had a further look

            I desoldered the 4901NF dual mosfets. There is a short on one of them between D1 and S1/D2 pins, D1 being the 12V input and S1/D2 being the coil - so obviously when this failed some voltage has gone down through the coil where it shouldn't go.

            There is a short circuit to ground an the other (load end) of the coil.

            I removed the 3V electrolytics and they are not short. There are 11 MLCC capacitors on this supply rail, that I can see though I may have missed some tiny ones.

            I decided to power up the very similar Phantom card to see what voltage is on this rail, but I then found on that card I have 12V supply but no Vcore/Vram. I will have a look at that but as there are no shorts, then for some reason the VRMs are not running at all.

            Going out on a limb here - as no-one has advised one way or the other I am guessing this shorted rail on the Palit is the DRAM supply.

            I can either put it on one side while I try fix the Phantom, to find out what the voltage rails should be so I can try trace the short with my bench supply at the correct voltage - or I can remove all the capacitors that I can see with a short across them to see if one of those has failed, or I can start removing the RAM chips to see if the short goes away. There is no visible damage that I noticed.

            I can get a pack of 10x 4901NF for around 5 euros and I could get 12 replacement RAM chips for about 15 euros. So if I am right this could be a 20 euro fix, which is definitely worth it. But if the RAM went short would that blow the GPU, even though the GPU resistance looks OK and Vcore would have been OK?

            This can't be the first card with this type of fault - so what usually fails in this scenario? Anyone know?

            cheers
            Last edited by dicky96; 04-20-2020, 07:37 AM.
            Follow me on YouTube
            ------------------
            Learn Electronics Repair
            https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

            Comment


              #7
              Re: GTX 780tI Repair attempt...

              While I am waiting see if anyone has any ideas how to proceed with the Palit I decided to have a proper look at the Phantom. AS I mentioned this one has no shorts that I have found, it has 12V supply but has no Vcore or Vram

              What I found is, the Vore has 8 phases and each phase is driven by a Renesas 20658BN



              On mine I have a circuit like the one in the datasheet '(2) VCIN 5V application' on page 7.
              I have 12V and 5V supply but no PWM input.

              Similarly I have no PWM drive to the gates of the two dual Mosfets that drive Vram

              I traced the Vcore PWM to a NCP4208 8-phase synchronous buck converter


              Here I found pin 1 (VCC3) has 3.3 - this is an internally generated power supply output, so I know the chip must have VCC power input, and is probably OK, but EN (pin 6) and PWRGD (pin 2) are low - so it is disabled.

              Now I had a bit of a think.... it would seem logical that this chip is enabled by some sort of processor or microcontroller when certain conditions are present. Also this chip has VOD0-VID7 inputs to set the Vocre. So something must control it.

              So I looked around the board and I could see there are two small inductors which means I must have two other power supply rails. One of these was giving 5V, the BIOS chip has 3.3V on pin 8 which looks sensible and suggest another regulator is working somewhere, but the other inductor has 0V on it. There is no short on this rail.

              I looked on the other side of the PCB to see what drives that one and this is when I found someone has been messing!!

              This is evidentally someone with some reasonable amount of knowledge of what they are doing but with very little evidence of good soldering skills.

              I found a few chips that I can't identify, so I don't know if one of them is a micro-controller - but I suspect the inductor with the 0V on it acts like some sort of standby or startup circuit and probably powers part of the GPU, which then in turn sets the VID0-7 on the NCP4208 to some default value, sets EN high and starts up the Vcore and Vram.

              At least I can imagine it is supposed to work like that and of course without the standby or startup supply present - nothing happens. So I need to figure out what is wrong with that circuit, and also clean up the other guys mess, but problem is I can't find a datasheet. What I am sure is the controller IC is marked 3A F2 F2G or 3A F2 F26

              Please see the attached pics. That will help. Does anyone recognise that chip?

              Cheers
              Attached Files
              Last edited by dicky96; 04-20-2020, 10:55 AM.
              Follow me on YouTube
              ------------------
              Learn Electronics Repair
              https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

              Comment


                #8
                Re: GTX 780tI Repair attempt...

                The part marked 3A could be a Richtek rt8071azqw

                https://www.eevblog.com/forum/repair...-component-id/
                Last edited by SMDFlea; 04-20-2020, 02:02 PM.
                All donations to badcaps are welcome, click on this link to donate. Thanks to all supporters

                Comment


                  #9
                  Re: GTX 780tI Repair attempt...

                  The one on the Palit is marked 3A FM L1G which makes sense

                  Yes it probably is that RT8071chip - it has 5V input on pins 2 & 3 which is correct. However pin 10, which would be EN, is low - so it isn't going to start.

                  https://www.richtek.com/Home/Product...web&sc_lang=en

                  So the question would be, where does the enable for this chip?
                  Follow me on YouTube
                  ------------------
                  Learn Electronics Repair
                  https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

                  Comment


                    #10
                    Re: GTX 780tI Repair attempt...

                    Originally posted by dicky96 View Post
                    I had a quick look at the other two cards with my multimeter. One of them (I think it was the Gainward Phantom) looks to be very similar to the Palit card
                    yes u may have already realized this but yes, palit and gainward both use the same oem basically. palit, gainward, xpertvision and manli all use the same oem for their stuff so amongst the same model of video cards, these brands are all the same thing.

                    so u can use this to compare and know what is working and not by examining the other card. momaka said its always good to get two of the same non-working cards so u can compare what it should be like when its working and not. that is very useful to figure out and reverse engineer the cards and get them working again. hope this helps u.

                    Comment


                      #11
                      Re: GTX 780tI Repair attempt...

                      Yes they are very similar

                      I think today I will remove all the RAM chips from the Palit as I read somewhere on a youtube comment that they go short if the high side FETs that drive Vram rail go short on the high side - which is what happened on my card. This confirms what I thought would happen

                      I already removed the Coils and faulty FET, so the card should power up anyway, if i link out the blown fuse for now. I watched a video on you tube where someone had removed all the RAM freom one of these cards but was still getting POST text appear on the screen when the PC boots.

                      That surprised me but unfortunately I don't understand russian so I have no idea what he was saying, I could only look at the pictures.

                      The Phantom, it's like Vcore and Vram are not enabled (though I didn't bother try find the EN signal to the Vram controller - I did find EN to Vcore controller and it is low).

                      The same applies to that other small buck regulator I found which Rasz on eevblog forum said was the 'PCIe controller PSU' and that the card would not boot up without it.

                      So my idea that the small converter should always be on, like a backup or startup rail, is clearly wrong.

                      So, what is the sequence supposed to be? The BIOS has power, but what enables the power rails? And is it the same chip that executes the BIOS rom code? There must be a micro-controller type device somewhere, surely?? There is just so much I don't know
                      Last edited by dicky96; 04-22-2020, 02:28 AM.
                      Follow me on YouTube
                      ------------------
                      Learn Electronics Repair
                      https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

                      Comment


                        #12
                        Re: GTX 780tI Repair attempt...

                        I'm getting deeper into this lol, I do hope this is entertaining some of you, at least.

                        I was trying to find how the GTX780 starts up, what enables the Vcore and Vram and that other supply rail I found

                        I had a good look around the PCB and found a little SOT23-5 chip near the PCIe connector which I assume is a LDO or something like that. It connects to VCC3 on Pin 7 of the PCIe connector.

                        I don't know what that is, it is marked VE and on the Gigabyte it is B8GO. I thought it may be this one
                        https://pdf1.alldatasheet.com/datash...P130K161D.html
                        which is available as a SOT23-5 but the pinout is wrong. On mine Pins 1 & 5 are connected together (I can see the PCB track with the chip removed), and pin 3 is ground, not pin 2.

                        On the Phantom board, this chip had a short circuit from Pin 1 and 5 to ground. On the Palit this reads 700 ohms.

                        I checked all the capacitors in that area (some of the caps in the areas circled in red on the pics read short) and any I could find that had a short, I removed and tested again - but the short is not caused by one of the caps, or the LDO which I also removed (unless I missed a cap somewhere)

                        I started thinking that this supply probably goes to the GPU. So I removed the GPU from the Gigabyte 780ti, which has a Vcore short - and yes the VCC3 supply from pin 7 PCIe does go to the GPU. So I think I found my 'startup' supply rail.

                        It looks on the Phantom I have a GPU that reads OK resistance on Vcore (and Vram resistance is good too) but has a short on the PCIe VCC3 supply, unless as I say I missed some capacitor or VCC3 goes somewhere else which has short. I guess the next thing to do is remove that GPU from the Phantom and see if the short goes away.

                        Now, very interesting, on the Gigabyte I still have a short on Vcore with the GPU removed! That came as a surprise!!

                        There is no short on the 12V side of the VRMs just on the Vcore side. Short circuit capacitor? None of them look distressed. But it obviously is not a shorted GPU

                        I think I may need to order a GK110-435-B1 stencil because it is possible the Gigabyte GPU is good......
                        Attached Files
                        Last edited by dicky96; 04-23-2020, 07:08 AM. Reason: corrected an error
                        Follow me on YouTube
                        ------------------
                        Learn Electronics Repair
                        https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

                        Comment


                          #13
                          Re: GTX 780tI Repair attempt...

                          Another day..... another update

                          PALIT GTX 780ti
                          This is the onewith a short on Vram but not on Vcore. I decided to power up the Palit with no Vram supply - the FETs and coils are removed. The GPU has 0.87V Vcore and gets nicely warm. I had a look at that VCC3 rail which is shorted on the Phantom (near the PCIe connector) and it should have 3.2V on it because that is what I get on the Palit.

                          The PC did not detect the card. I saw some youtube video where an engineer has powered up a GTX 780 with all the RAM chips removed - and he gets a BIOS POST screen with text. I don't know how that works with no RAM fitted - the video was in Russian :-( This board has a Vram short - either I need to remove all the RAM or look for shorted caps.

                          GIGABYTE GTX 780ti
                          This is the one where I removed the GPU and still have a short on Vcore. I couldn't see any difference in resistance to ground on either side of the Vcore coils using my multimeter so I used my ESR meter to check the resistance from the coils on Vcore to ground on the supply and load side. Obviously the inductor acts as a serial resistance at the frequency the ESR meter uses - so it is obvious which side of the inductor has the short!

                          I could see there was 0.4 ohms on the load side and 0.7 ohms on the supply side so I removed all the electrolytics by the coils on the load side but it didn't make any difference. Some of these are regular SMD electrolytics and some are flat black ones. I managed to break one of the black ones. What are these type called? See pics.

                          I then had a look around under the microscope in the same area and found a distressed looking MLCC which read short. I removed this and it was short, but the short was still present on Vcore. It just changed from 0.5 ohms to 1 ohm

                          I then did a google and found these GPU should take up to 1.2V. So I used my bench supply to send 1.2V into the short - actually I had to use my voltmeter to monitor the voltage at the point I soldered the wires to Vcore supply and turn my PSU up to 2.1V to get 1.2V on the board due to the resistance of my leads. I then had about 1 amp flowing and found a very hot MLCC on the opposite side of the board under where the GPU would fit! Remove that and hey presto the Vcore short has gone!

                          So if I order stencil for the GPU and replace the two shorted MLCC plus the other black capacitor I broke then maybe - maybe - maybe this one will work. Dependent on my skills to reball and refit the GPU, amongst other things. Maybe I didn't need to remove the GPU at all but I am not sure I would have found the two short MLCC caps without doing that.

                          PHANTOM
                          This is the one with a short on VCC3 supply from the PCIe connector
                          I was just going to use my bench supply to send 3.2V into that supply (as per the the voltage I measured on the Palit) when I noticed a bit of a mess in a similar area to where I found the shorted Vcore cap on the Gigabyte. It looked like some spare solder blobs and a probable solder bridge. I cleaned that up with flux and braid and the Short on VCC3 had gone I didn't try replace the six missing components I just powered it up - and it still does not detect. Then I found I still have no VCC3 I can compare with the Palit to work out what components are missing.

                          A quick check to see why I still have no VCC3 after removing the short - I found on the Palit where I do have VCC3 I have 0 ohms from Pin 7 PCIe to that voltage rail. But on the Phantom I have 47 ohms. I couldn't see that before because of the short.

                          This is now looking like I have an open circuit fuse or a fusible resistor (or just a low value resistor gone high) somewhere connected between pin 7 PCIe and that voltage rail!!

                          However at this point I had well and truly had enough of these hateful graphics cards! Why are they giving me so much hassle Fix one thing and then it's another bloody thing aarrrghh!

                          So I will look tomorrow when I calmed down. Could be this one will be fixed.
                          Attached Files
                          Last edited by dicky96; 04-24-2020, 10:30 AM.
                          Follow me on YouTube
                          ------------------
                          Learn Electronics Repair
                          https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

                          Comment


                            #14
                            Re: GTX 780tI Repair attempt...

                            Originally posted by dicky96 View Post
                            However at this point I had well and truly had enough of these hateful graphics cards! Why are they giving me so much hassle Fix one thing and then it's another bloody thing aarrrghh!
                            Hi dicky!

                            I've been reading this thread on and off during breaks at work and meaning to reply to your posts for a while. Before anything else, first I have to say, my hat off to you for getting his far. You're doing absolutely great, so don't give up! Video cards can indeed be riddled with pesky problems and in fact aren't always repairable. But if you take your time and try not rush through them, you'll be able to get down to the problem with most of them. That said, I don't know if repairing video cards for resale can be worth it or not. But for me, at least, I do it more for fun / entertainment and learning purposes than anything else - and those, too, have value.

                            Originally posted by dicky96 View Post
                            Going out on a limb here - as no-one has advised one way or the other I am guessing this shorted rail on the Palit is the DRAM supply.
                            Correct.

                            In fact, it's not just the RAM Vdd supply, but also part of the GPU uses it (I think so it can communicate with the RAM.) Moreover, I think the GPU actually draws a lot more power from that supply than the RAM itself (but of course, less than its main GPU V_core supply.) The RAM chips actually draw very little power and probably could do away with a series MOSFET linear power supply (and on older DDR PC motherboards, they often did.)

                            On that note, it was actually common for the GPU chips on the older Radeon HD series of video cards to develop a short-circuit on the RAM Vdd supply (due to inadequate cooling and/or prolonged running of the GPU at high temperature and its substrate BGA failing.) This usually resulted in either 1) a MOSFET on the RAM Vdd VRM dying and causing a short-circuit for the PC PSU (making the PC not start) or 2) RAM Vdd VRM MOSFETs running very hot in current-limiting mode (more rare, but happened to a number of Xbox 360's.)

                            Originally posted by dicky96 View Post
                            or I can start removing the RAM chips to see if the short goes away.
                            I have not seen RAM chips go bad by themselves to date. Even when an "upper" MOSFET (one between 12V supply and VRM output) shorted, most synchronous VRMs will just ramp up or completely turn On the "lower" side MOSFETs to keep the output from going over-voltage, thus protecting the RAM chip from damage. The only time I actually had RAM chips go short-circuit was on two EVGA GeForce 7600 GT cards with those dreaded Sacon FZ caps. The FZ caps had failed so badly that there was nothing to filter the VRM outputs. As a result, the RAM VRM sent massive voltage spikes through the RAM chips and killed them... then burned the RAM supply inductor.

                            Originally posted by dicky96 View Post
                            This can't be the first card with this type of fault - so what usually fails in this scenario? Anyone know?
                            I would put my money either on a shorted ceramic cap on the RAM Vdd supply or the GPU chip actually shorted on the RAM supply.

                            To find out what it is, indeed you either need to remove all of the RAM chips or all of the MLCCs or the GPU chip itself. In that regard, I think it's best to remove whichever one of these you think would be the least amount of work and then go from there. I personally would pick removing the GPU chip, as then you can use your bench lab power supply to try and determine where the short circuit is. With the GPU in there, your bench power supply won't have enough power to do that, as the GPU will suck up all its power.

                            Originally posted by dicky96 View Post
                            I found a few chips that I can't identify, so I don't know if one of them is a micro-controller - but I suspect the inductor with the 0V on it acts like some sort of standby or startup circuit and probably powers part of the GPU, which then in turn sets the VID0-7 on the NCP4208 to some default value, sets EN high and starts up the Vcore and Vram.

                            At least I can imagine it is supposed to work like that and of course without the standby or startup supply present - nothing happens.
                            You got it!

                            The GPU chip indeed has a "standby" supply or supplies (can be more than one, if I recall)... actually, it's probably more accurate to call it a "startup" supply, as this is typically only used by startup logic and maybe very basic video processing. After all, remember the Radeon 9700 from back in the day? This was the first card I encountered with a separate power connector. I remember forgetting to plug in the floppy power cable from the PSU one time, and the card still booted up with a red text on the screen, saying there is an Error and the power cable is not plugged. The fact that the card could display something on the screen in basic text mode itself suggested that it had multiple GPU power rails/supplies. A lot has changed since then on modern GPUs, but the basics are still the same more or less.

                            Originally posted by ChaosLegionnaire View Post
                            so u can use this to compare and know what is working and not by examining the other card. momaka said its always good to get two of the same non-working cards so u can compare what it should be like when its working and not.
                            Yes.

                            Not only that, but I find that a lot of people don't store their PC hardware properly - especially video cards, it seems. As such, I've gotten more cards with knocked off SMD parts now that I have without any damage - even when I bought supposedly "tested and working" used video cards.

                            Thus, if a video card doesn't come with any metal back plates that completely covers all of the components, it's always a good idea to run a thorough inspection and looked for chipped or missing SMD parts.

                            Normally, if it's just a few ceramic caps, that's no big deal. Most ceramic caps do filtering, anyways, and in many cases the card will be able to run without. But not all are like that. And even worse, if a config SMD resistor or logic transistor is knocked off - finding out what it was can be a hassle. So for these, finding out what they are will be a lot easier if you have the same card to look up the parts or take them off and measure the values.

                            Originally posted by dicky96 View Post
                            GIGABYTE GTX 780ti...
                            I then did a google and found these GPU should take up to 1.2V. So I used my bench supply to send 1.2V into the short - actually I had to use my voltmeter to monitor the voltage at the point I soldered the wires to Vcore supply and turn my PSU up to 2.1V to get 1.2V on the board due to the resistance of my leads. I then had about 1 amp flowing and found a very hot MLCC on the opposite side of the board under where the GPU would fit! Remove that and hey presto the Vcore short has gone!
                            Nice work there!
                            That Gigabyte indeed sounds like the most promising card right now to be working again... granted the procedure with soldering back the GPU goes well.

                            Originally posted by dicky96 View Post
                            Maybe I didn't need to remove the GPU at all but I am not sure I would have found the two short MLCC caps without doing that.
                            No, unfortunately, you would NOT have found the short if you didn't remove the GPU. Reason why is because the GPU chip has very low resistance. Trying to feed any voltage on the GPU V_core rail with the GPU still in there would have made the GPU suck up all the power - more than your bench supply can provide, anyways. Thus, you wouldn't have been able to get a high enough voltage to detect which MLCC is shorted/bad.

                            So you did the right thigh with removing the GPU. It's just... a lot of work for finding a short. :\ But that's just the way it is.

                            Anyways, keep us posted on what you find.
                            And at the very least, any work you do on these GPUs will surely polish your troubleshooting electronics skills, if nothing else.
                            Last edited by momaka; 04-24-2020, 10:48 PM.

                            Comment


                              #15
                              Re: GTX 780tI Repair attempt...

                              Hi,
                              Thanks for your post. Any progress?

                              I have a Nvidia Quadro K5000 GPU. It is working without driver (with default Microsoft driver), I can turn on the computer, but when I install the right driver, the fans starts spinning with 100% load and in the device manager I see a yellow (!). I tried with linux too, same result.
                              I have very little knowledge of electronics, but I took it apart and I found many shorted resistor under (other side) the GPU chip. I think it can't be good sign.
                              What do you think? the GPU chip is faulty? Or can be anything else? What should I check?
                              I really need this GPU for my work and I can't buy a new one right now.
                              Thanks for your help.
                              Last edited by piernov; 10-22-2022, 08:31 AM.

                              Comment


                                #16
                                Re: GTX 780tI Repair attempt...

                                Hi mateksz,

                                Welcome to the forums!

                                In my experience, video cards that work OK with the Windows standard VGA drivers but crash when the proper drivers are installed is typically due to a faulty/dieing GPU.

                                As for the "shorted resistor"... by that I assume you mean MLCC / SMD ceramic caps. They can be shorted, but it's not very common. And even if there is nothing wrong with the card, you will find that these still measure very low resistance. Many people mistakenly think they are shorted when they are not. Take a working video card and measure those caps and you will see that they will also show up as "short circuit". This is normal, unless the resistance really is showing as 0 Ohms. It's not uncommon for GPUs to measure down to 1.5 Ohms on their V_core supply rail (especially if the GPU chip is still warm after testing.)
                                Last edited by piernov; 10-22-2022, 08:32 AM.

                                Comment


                                  #17
                                  Re: GTX 780tI Repair attempt...

                                  I've said that many many times but the point is to *not* melt solder. Melting solder can cause more problems than there is already. Most GPU will react to heat at below 200°C. You don't need to melt the solder balls when the solder balls are not the problem. That's why baking a board in an oven other silly tricks work for a lot of people while there's no way they are actually reflowing the chip properly.
                                  Bumpgate is for 2006-2008 NVidia chips. There's a bit of literature on semiaccurate about this. But other chips can fail too (well known failing Radeon HD 6000M or RS880), however there's no study on them, so bump issue or something else who knows. Fact is that they do also react to heat, 200°C will resurrect them for a short while. In any case a dead GPU is a dead GPU.
                                  As for contact issue between the chip and the board, yes they do happen, but it's pretty rare , and even rarer when it's the ball themselves, especially when there's not much external mechanical stress. By experience when you have a chip with contact issue (you press down on it and behaviour changes), when you pull it you'll usually have some pads coming off as well, not because your profile is bad (if you poke the chip and it moves easily not getting stuck in one corner the balls are melted), but because the pads were damaged to begin with. Reflowing a chip won't fix that.
                                  I've also seen some people argue that the thermal dilatation of the balls is what fixes them temporarily. Well ok I must admit it doesn't sound like a completely unreasonable theory when you resurrect a chip for a few hours. But I really doubt it when chips keeps working for days/weeks with several thermal cycles. Also why would some chip be affected more than others if the issue isn't the chip itself.
                                  I also admit that I don't have a definitive explanation as to what the failure mode is and why heat actually revives them for a short while (apart from NVidia bumpgate which has been explained). I'd love to have one, but I'm still convinced it's not the solder balls.
                                  OpenBoardView — https://github.com/OpenBoardView/OpenBoardView

                                  Comment


                                    #18
                                    Re: GTX 780tI Repair attempt...

                                    Originally posted by momaka View Post
                                    Hi dicky!

                                    I've been reading this thread on and off during breaks at work and meaning to reply to your posts for a while. Before anything else, first I have to say, my hat off to you for getting his far. You're doing absolutely great, so don't give up!
                                    You know what mate, I really appreciate that comment, and coming from someone on here it means a lot to me. Thanks!

                                    So an update as was asked

                                    GIGABYTE
                                    The Gigabyte 780ti I have on one side - waiting a stencil for the GPU and some capacitors. Guess they will take a while to arrive here.

                                    PHANTOM
                                    The Phantom GTX 780 I've had something of a result. Despite spending a couple of hours on this, I just could not find why I have 25 ohms (was 47 ohms but it changed for some reason) between pin 7 of PCIe (VCC3) and the voltage rail it should supply on the graphics card. I have a direct connection from PCIi pin 7 to one capacitor on the graphics card but to everywhere else I can see it goes, I have 25 ohms.

                                    As hard as I looked I can't find a resistor or similar that accounts for the resistance. My best guess is a faulty via causing some resistance???

                                    So I decided to cobble it. I soldered a wire from the MLCC which was connecting to pin 7 PCIe and soldered the other end to another capacitor on the 3.3V voltage rail on the graphics card - thus 'bypassing' (or shorting out) the 25 ohms.

                                    I then powered up the card and could tell straight away (because the GPU now warmed up) that I have Vcore!! But the graphics card was still not detecting I was getting the usual beeep bip bip bip same as before That was enough for yesterday.

                                    Anyway it occurred to me later that really by now I should be getting a picture but maybe the old MSI motherboard I was using as a test rig (MS-7293 which came out of an old Fujitsu tower server PC) was not compatible with the graphics card - like the BIOS couldn't detect it or something. This despite the same test motherboard working OK with a Geforce 7600GS I have.

                                    The reason for this thought is because I have had problems in the past with machines like Dell and HP not detecting some graphics card when you fit them.

                                    I had used that motherboard because I had it lying around and it does not have onboard graphics. So anyway I got another old board I have that also has no onboard graphics, an Asus P5B Deluxe.

                                    I put the Phantom in that - and it detects!!! BUT..... although I now have Vcore (0.94V) and Vram (1.5V) and I get a picture - there are little white dashed lines on it See attached pics.

                                    Only thing I am thinking now is maybe the six little SMDs on the rear behind the GPU that were missing and/or someone had bodged up with solder (see pic) are actually necessary and causing the messed up video output. I could take these from the Palit and see what the parts should be.

                                    Or maybe the GPU or RAM are screwed up and I am

                                    Palit GTX780ti
                                    I tried with this one on the Asus P5B a swell. It still does not detect. I have Vcore 0.87V. I currently have no Vram as the inductors and the short circuit FET are removed. I believe a GTX780ti card should still detect and you should get a picture of sorts if the GPU is active but there is no RAM fitted on the card. Is that correct?

                                    I think the next thing on this one is to remove either the GPU, RAM or both to try find the short on Vram. Vcore resistance reads OK.

                                    Gigabyte GTX770
                                    I've still not really looked at this one properly, it has a blown VRM FET (with a hole in it) but no short on Vcore or Vram.
                                    Attached Files
                                    Last edited by dicky96; 04-28-2020, 06:39 AM.
                                    Follow me on YouTube
                                    ------------------
                                    Learn Electronics Repair
                                    https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

                                    Comment


                                      #19
                                      Re: GTX 780tI Repair attempt...

                                      A little more on the Palit.

                                      I tried pushing 1.5V into Vram on the Palit using my bench supply, seeing as that is the Vram voltage I can read on the kinda working Phantom. I had it sitting at a steady 3 Amps for several minutes but nothing gets warm.

                                      I want to try removing the Vram chips next. There are 12 of them. I can either use my semi auto BGA station but would require numerous heat cycles but at least it is a controller profile (I can put a small upper heat nozzle over each RAM chip in turn). Or I could do it manually with my KADA 853B preheater under each RAM chip and my Quick 861DW. I have a manual vacuum pick up tool also, Aoyue 932.

                                      Recommendations?

                                      I've never tried this before apart from stripping parts from old boards where i don't really care about the board.

                                      Rich
                                      Follow me on YouTube
                                      ------------------
                                      Learn Electronics Repair
                                      https://www.youtube.com/channel/UCFX...R8UZ2vg/videos

                                      Comment


                                        #20
                                        Re: GTX 780tI Repair attempt...

                                        That KADA 853B preheater is a small one?
                                        Use your BGA rework station as preheater. Videocards have lots of copper sucking away heat. Use the Quick to remove the ram chips.

                                        Comment

                                        Working...
                                        X