Graphics Card Autopsy - MSI GeForce GTX 980 Ti Golden Edition

[@DayLateGamerWill]

Makes me happy to hear the start of this video, I *still* own and I'm still using my MSI 980ti Gold Edition. Surprisingly good lil work horse of a card, picked it up in 2016 and have been gaming on it ever since! Just upgraded to a i7 12700k, 32GB RAM, and the venerable 980ti Gold Edition, and it will hopefully carry me all the way to a 40 series RTX card ^_^

[@HazewinDog]

I just got two of these in a bundle for €350! They came with MSI-branded EK waterblocks attached, so I came here to figure out how to re-attach the stock coolers that were also supplied. The shots were a bit too cinematic, but enough to help me figure it out regardless :) Thanks! Also, I subscribed. Judging by this video, your channel seems right up my alley :) I enjoyed the troubleshooting, and really appreciated the advice at the end. I had no idea that could kill a GPU!

[@lepesteper]

I'm learning so much from you. Thanks a lot

[@DaKosi]

Very interesting that supplying too little voltage can damage the card too. Good video. Can't repair them all.

[@sebastiandenenfeld9886]

Amazing explanation,you are great teacher!

[@مغاربةاحرار-غ2ش]

Thank you for the valuable scientific information that you shared with us, and if you would kindly ask us, we want more if it is possible to examine it in another way to spread the benefit and learn the various methods used to examine this type. Regardless of the result and think it remains one through your evaluation, thank you very much.

[@kresimircosic9035]

Cool clip, wish I had the know how and equipment like you do to do stuff like this.

[@pjprodukt]

Must be what happened to mine 2 months ago. Funny how I came here to search if anyone had a similar issue. Tried reapplying some thermal paste, but that did nothing. Didn't dive as deep as you did. Worst time to buy a GPU right now too...sigh.

[@AaronAverett]

Oof, yeah, that’s a tough break. If the board didn’t delaminate like mine did, it’s probably not beyond repair, but you’d probably need to replace all eight of the nfet ICs to get it to work reliably.

[@ataxela5208]

Hi, I had the same MSI GTX980Ti model, I came looking for an answer because my PC abruptly shut down in the middle of gaming, and there's smoke coming from my 5-year-old GPU. This is the EXACT problem with the EXACT model. Instead of searching aimlessly through Google, I came here. Thanks for the video, I can close my case and move on at the same time... I'm so pissed off that my GPU broke in the middle of a pandemic and shortage supply going on the GPUs.

[@AaronAverett]

Did you pull the heatsink off and inspect the board? You sometimes get lucky, and find that the board itself wasn't damaged. In those cases, this is a very repairable problem for someone with the right equipment.

[@aysat]

Well, one way to check were the short took place is to remove that gpu core chokes and measure from each side. i am sure if you manage to fix short in that burnt area ,card will run just fine. other way is to remove the gpu chip and measure.

[@AaronAverett]

You are correct. Pulling the chokes off and hunting for shorted components with a power supply is a totally valid tactic, but I think we know where the short is in this case, and the problem is just removing enough of the damaged board to clear it without destroying any really critical traces. I actually considered making a follow up where I'd do that, but I just haven't had the time. Removing the GPU itself is a totally different proposition. You can't just remove it with hot air, and once you get it off, it's a pretty involved process to reinstall it.

[@aysat]

Reballing is a headache actually :) but i am expecting for you to remove the short and see picture on screen !

[@مغاربةاحرار-غ2ش]

Thank you for the valuable scientific information that you shared with us, and if you would kindly ask us, we want more if it is possible to examine it in another way to spread the benefit and learn the various methods used to examine this type. Regardless of the result and think it remains one through your evaluation, thank you very much

[@calin1109]

Hello, thank you for sharing your knowledge with us. I'm having trouble with my gtx 980Ti and i was curious if you've ever meet this problem with other gpu's. Mine isn't brick dead, it has video signal but it restarts my pc or it freezes it even if the pc is at idle or in full load. I have put my gpu in another pc and the behaviour is the same. Some times it restarts the pc even if it is at idle sometimes i play a game and it's ok an sometimes i open a game and it crashes. Can you please give me an advice on what to check? Thank you!

[@AaronAverett]

That sort of behavior is generally the result of a BGA failure, most likely at the GPU itself. Unfortunately, that is not something I'd describe to most people as repairable.

[@calin1109]

@@AaronAverett Thanx for answering. It seems that I'll have to wait for the prices to drop ang get a new gpu.

[@TomerDlv]

hello I have an EVGA 980 ti scc never been overclocked, PC powers up fine, GPU works fine on normal work (internet brawsing, youtube etc...) but the moment i put it under load(lanch a game) , it crashes, black screen (sometimes white for a second), GPU light flash and turn off for a second than power on( like its doing a restart) , but no video afterwards. i have to restart The PC to get it to work again. *fans are working fine any suggestion?

[@AaronAverett]

If you have the right tools, you could try measuring the voltage at each of the logic power outputs to see if one of them is shutting off, which is a remote possibility. Most likely, though, your problem is a BGA failure, which is a very difficult thing to repair, even for someone with the tools and experience to do it.

[@TomerDlv]

@@AaronAverett thanks for the answer, really appreciate it!

[@mr.k3224]

Some advice would be highly appreciated. Since i dont know on which forum i can find you ill present my problem here. I have faulty gtx 970 4gb gainward phantom. Fans spins but no post or picture. As far as i can tell all voltages are present, but there is no voltage on vcore or ram after power rail. No voltage on gates either. On vcore power rail is ncp5369 i dont know where the gate is. Vcore resistance is 4.8 ohms and ram is around 90ohms. Should i remove coils and check if power rail is faulty or anyone has better idea? Thank you in advance!

[@AaronAverett]

You have to measure output of each VRM at the output side of the coils. If you're measuring at the output of the coils for the core and memory, and you have zero volts to ground at both of them, then clearly, not all of your voltages are present. I'm just guessing here, but I if you're missing both the memory and core voltage, you should probably start your troubleshooting with the memory rail, I think I talked about this in another video, but the power rails all start up in a specific order, with each one eventually turning on the enable signal for the next one in the sequence. The card's I've looked at usually start the memory rail up first, so if you're missing both, your problem is likely the one that earliest in the sequence. If you look at the data sheets for the VRM controller ICs, you should see that one of the pins is an "enable" signal. In addition to the supply voltage, the IC requires that you supply a voltage on that pin in order to run. Try checking all of those and see if you actually have a voltage on all of them.

[@LawrenceTimme]

Rip. Such a nice card

[@idontwatchtv]

I've been watching your videos in hoping to save my old HD 7990 that's been tucked away for quite sometime. I wanted to retest it to make sure it works as I plan to sell it. It used to work without issues back in 2015 but when I tried to use it recently there was a short spark when I turned on my PC to what seems to be the same issue with that 980 Ti GE. I used an EVGA 850W G2 PSU which should definitely be more than enough for my HD 7990, the PSU I used back in 2015 was an XFX 650W Gold XTR. I don't think it's my EVGA PSU that caused it as my OC'd 2080 Ti runs. I plan to get a multimeter just to find out what could've happened with it but do you think it's salvageable? imgur.com/a/Q7KuoL8

[@AaronAverett]

Holy crap, that's gnarly! It's hard to say whether it's repairable without having the card in hand, but I suspect not. Usually, when you see damage like this where the board itself delaminated, the layers inside the board are shorted together such that you can't really clear the short without grinding away a significant portion of the board. That said, if you have the equipment, it's worth a shot. It looks to me like the root cause of the failure is that cracked MLC cap (C532). If you remove it, and you don't have a short across its terminals, there's a reasonable chance the card may still work if you can replace the other couple of caps that were in that area.

[@idontwatchtv]

@@AaronAverett Thanks for the reply. I'm no electrical engineer nor have a full on background on it so I'm just going by KZbin Videos on how to do such things. I went looking around at digikeys but kind of confused on what I'm looking for. I'm looking for a C532 Surface Mount Chip Resistor right? www.digikey.com.au/products/en/resistors/chip-resistor-surface-mount/52?k=c532 I don't have a hot air station or soldering station nor any experience on using them but if I can manage to narrow down some idea on what happened to my HD 7990 I might go ahead and get those to see what I can do.

[@AaronAverett]

The markings on the board are not values, but names for each component that you could use to look up the value in a schematic, if you had one. Unfortunately, the schematics for most graphics cards aren't available to us as consumers, so we have to use other means. C532 is a multilayer ceramic capacitor, which we know based on its appearance and based on its beginning with a C, which by convention normally indicates a capacitor. To find its value, we'll need to remove the corresponding capacitor from one of the other phases and measure it with a multimeter in capacitance mode. We'd then reinstall the removed capacitor, and order a replacement for our damaged one. Don't forget that you still need to make sure the short is cleared by removing the failed capacitor. If it's not, you'll likely need to scrape away the damaged traces in that area until it is. You'll also need to replace C524, and the two smaller caps below it, which will likely mean running a jumper wire from somewhere. The hot air station I used in the video is an Aoyue 968A+. I like it - it's good enough for a hobbyist, but I'd probably want a nicer one if I were doing this for a living. You'll need flux and solder as well, which don't come with it. This video discusses a problem almost exactly like what you're looking at, but my card had fuses to keep the short from delaminating part of the board. There's a part 2 where I actually get it working. kzbin.info/www/bejne/g4SZi5aYeLl-p9E

[@idontwatchtv]

@@AaronAverett Seems like quite a work ahead of me but nonetheless it's a good way to learn more about the Electronics of Graphics Cards. I'm still awaiting for the multimeter but will be a look out on Hot Air Stations once I figure out my HD 7990. I have actually watched that video you linked but will most likely look at it again for reference guide in the future. As for the replacement capacitor, what values or indicators do I need to specifically look/type into so that I order the correct one? Thanks for your responses, greatly appreciate it.

[@AaronAverett]

@@idontwatchtv The caps will have a number of different specifications. The most important are capacitance and physical size. Capacitance is measured in Farads (F), with the cap you're looking at there most likely being something like 10 uF (that is, microfarads). The physical size is obviously measured millimeters, but SMD components are often standard sizes, which are represented by a 4 digit number. C524 looks like an 0805 to me. C352 looks like a 1210 or 1206. As long as it fits on the pads and meets the capacitance and voltage specs required, the actual size doesn't matter, but you can look up the SMD package size codes and then measure them in order to make sure you're getting the right size. The voltage also matters, but it's more of an absolute maximum rating than something you have to match exactly. I suspect these caps are on a 12V circuit (the input from the power supply), so anything over that should work, but I'd look for 25V rated caps if I were going to the trouble of replacing them.