Announcement

Collapse
No announcement yet.

EVGA Titan X Maxwell suspected bad caps

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

    EVGA Titan X Maxwell suspected bad caps

    Hello guys!

    I have an EVGA Geforce Titan X (Maxwell) with suspected bad caps. The card boots fine into BIOS and Safe Mode, as well as regular Windows without drivers (VGA mode). As soon as the drivers kick in the PC blue screens (actually black screen, but no display output during the blue screen). Also, I get a small number of artifacts in VGA mode in any resolution higher than 640x480 but none at that resolution.
    I've attempted to downclock the card by doing a card bios flash through nvflash as well as trying a stock bios flash, no results. Thermal paste and pads are in order, other than that no hardware fixes have been attempted. Card is not shorting to backplate either, tested without backplate with same results.

    The caps pictured are very suspect and seem to have bulged, assuming caps on the board are Lelon. From a different thread I surmise that the caps might be bad and should be replaced, and they did not speak too fondly about Lelon caps in this thread. ( https://www.badcaps.net/forum/showthread.php?t=24377 ).

    Caps read: F5 OCRZ 270 16v

    Thinking of ordering these to replace them, as they were suggested as a replacement in the thread I linked before.
    https://www.digikey.co.uk/product-de...018-ND/1136108

    Can anyone think of anything else to look out for before I replace the caps? Thanks in advance for any advice you can give.
    Attached Files

    #2
    Re: EVGA Titan X Maxwell suspected bad caps

    Stupid question: Card is out of warranty, like my ol' EVGA GTX 970 SC ACX2.0, I assume?

    Bad caps are of course possible, but from my experience, problems with GPUs, especially when they artifact are more often with weak soldering spots on the BGA, means it would require a reball/reflow.

    There are, in theory, ways to semi "reflow" a GPU with a heat-gun or even reflowing in an oven. I am looking into this myself, I just want to read up more on "reflowing" with a heat gun..which is a little risky when you can't measure temperatures. On the other hand, if the card is already defective and it is OOW, you don't really have anything to lose?

    Comment


      #3
      Re: EVGA Titan X Maxwell suspected bad caps

      I'm worried it might be a GPU solder issue or a bad memory module. However since they appear to be working in VGA mode I figured there's not much to lose by testing all the mosfets and caps. One thing I've noticed is a 2.5ohm short to ground on all the solid caps behind the chokes, but since the mosfets appear to be OK and I have no schematics available, I can't begin to figure out where the apparent short is coming from, and the card boots so I'm not sure what to make of it.

      The card is out of warranty and I bought it as faulty on ebay to repair, partly for fun and if it works I'll slot it in my pc for a while.

      Artifacts are very few and do not appear in low resolutions, but on higher resolutions they become more visible. This makes me think it might be a memory issue. But a memory issue could also be a power delivery issue I'm guessing, so it makes sense to replace these swollen caps.

      Comment


        #4
        Re: EVGA Titan X Maxwell suspected bad caps

        @flexy123
        You don't reflow a GPU. A GPU with artifacts or crashing with drivers is a dead GPU. Except when you have power supply or VRAM issue but that's very rare compared to the amount of dead GPUs.
        Only a few select boards were having soldering issue, which makes up for not even 0.01% of total GPU failure I guess.
        In almost every case the chip itself is defective, you may want to read on NVidia bumpgate which was affecting 2006-2008 GPUs, but later ones fail as well, especially AMD (it's probably a different issue than what NVidia had though, but it's still the chip itself).

        @Vilbergo
        You can try replacing the caps but I wouldn't bet much on it.
        2.5ohms is not a short, since it's GPU VCore this low resistance is perfectly normal.
        If you really had a short the card wouldn't work at all and most likely the PSU would shut down.

        Also never buy a dead graphics card or a used high-end one on eBay (as well as any laptop/computer that can have a failed GPU such as anything with 2006-2008 NVidia, 5000-6000 series AMD and the like) . They (almost) all have dead GPU. Sellers even bake them to make them work for a week or two.
        Yes baking them will often resurrect them for a short time. That's not a fix and it will often fail again in a couple of weeks. This is not a reflow, when baking the board inside an oven you set the oven temperature to 200°C which is not enough to reflow lead-free solder.
        Last edited by piernov; 08-05-2018, 05:05 AM.
        OpenBoardView — https://github.com/OpenBoardView/OpenBoardView

        Comment


          #5
          Re: EVGA Titan X Maxwell suspected bad caps

          I have tried to reball a Titan Black once, it didn't do anything different after reballing, it would still crash when drivers were installed. The GPU chip is dead, sorry.
          Originally posted by PeteS in CA
          Remember that by the time consequences of a short-sighted decision are experienced, the idiot who made the bad decision may have already been promoted or moved on to a better job at another company.
          A working TV? How boring!

          Comment


            #6
            Re: EVGA Titan X Maxwell suspected bad caps

            If the card was baked at some point it would explain why the caps and plastic connectors look like they've been subjected to such high heat. Thanks a lot for your comments, I'm not bothered about losing my initial investment but I'll chuck it back on eBay and see if I can recoup some of it and mark the card as for-spares.

            Comment


              #7
              Re: EVGA Titan X Maxwell suspected bad caps

              Originally posted by Vilbergo View Post
              The caps pictured are very suspect and seem to have bulged, assuming caps on the board are Lelon.
              ...
              Caps read: F5 OCRZ 270 16v
              Those are solid polymer capacitors. Unlike electrolytic caps, these don't bulge when they fail. The only time they can bulge is if they are subjected to extremely high heat - i.e. someone tried to reflow the video card in the oven and either set the temperatures too high or left the card in there for too long (or both ).

              Which bring me to....
              Originally posted by Vilbergo
              The card is out of warranty and I bought it as faulty on ebay to repair, partly for fun and if it works I'll slot it in my pc for a while.
              Buying used broken ("for parts or repair") cards on eBay is always risky business (actually, that's an understatement) and there's always a chance someone tried baking the card to see if it works before chucking it on eBay. I find that very few sellers are honest enough to say that they did this - at least when it comes to individual sellers. Recyclers are different, as some of them just don't have enough time to test all of their stuff and just list the cards for parts due to not being tested. I've bought from sellers like that before and got a perfectly working card.

              Originally posted by piernov View Post
              A GPU with artifacts or crashing with drivers is a dead GPU... In almost every case the chip itself is defective, you may want to read on NVidia bumpgate which was affecting 2006-2008 GPUs, but later ones fail as well, especially AMD (it's probably a different issue than what NVidia had though, but it's still the chip itself).
              ...
              Yes baking them will often resurrect them for a short time. That's not a fix and it will often fail again in a couple of weeks.
              I agree.

              In general, the GPU needs to reach about 215-220C for the solder between the die and substrate actually starts to melt. And due to the die having epoxy underfill, there's never a guarantee that a "reflow" (be it oven bake type or heatgun type or even proper BGA machine) would fix the GPU.

              That said, I do find that some GPUs will "reflow" okay and work for quite a while, especially if care is taken that the reflowed GPU does not overheat again (which is exactly what caused it to fail in the first place). This means keeping the GPU as cool as possible - below 60C at all times. For ATI/AMD cards like the HD4k, HD5k, and HD6k, you want to keep that temperature under 50C - and that's when the card is under full load. Most stock heatsinks do not have that kind of cooling capacity, so you either need to upgrade the heatsink (which often does not make sense to do due to costs) or switch to liquid cooling (which again can be almost as pricey as getting a new GPU). For this reason, I usually stick to mid-range and lower tier "high-end" cards, as finding a bigger heatsink for these from a higher-end model is the easiest and cheapest way to upgrade the cooling. Or if the card is under 120-130 Watts TDP, I find that sticking a big CPU cooler on it (with a ghetto mod) often does the trick to keep the card cool - that is, if you don't mind blocking all your expansion slots below the card.

              I personally have "reflowed" many different video cards. But from all of these, only the HD2k-4k and the old Radeon 9700/9800 are still working fine after several years (though I will also note I haven't used them that much). On the nVidia side of things, I can't seem to get a reliable reflow on anything - at least from their video cards. The GeForce 9x00 are the most promising from what I read, though (at least for nVidia). Anything with GTX in the name I stay away from. On the chipset-side of things, GeForce 6150/6200 chipsets do reflow okay most of the time, even at lower temperatures. The key to keeping them working for longer time is, of course, keeping them cool, as stated above. My reflowed Compaq Presario v6000 is still working after almost 2 years ago when I reflowed it (though it recently lost its wireless, which means chipset issues may be coming back again). Meanwhile, a GeForce 6150 chipset on a desktop PC I had that would often not detect the ethernet controller on a cold boot is now okay after just upgrading the cooling (adding a fan to the chipset heatsink). So despite the GeForce 6150/6200 being one of the most notorious chipsets for failures, they do actually "reflow" okay most of the time. As for how long they will work - that depends a lot on the cooling you put on them afterwards and also a bit of luck.

              But all in all, an artifacting GPU is a dead GPU and will eventually die again. In some cases, a reflow will just buy you a bit of time (or a bit more with good cooling and some luck).
              Last edited by momaka; 08-21-2018, 01:01 AM.

              Comment

              Working...
              X