What happens when capacitors go bad on an EVGA 3090 ?

Added: 2026-05-11

[00:00:00]Hello internet. This is a very interesting 390 EVJ that I have never seen before to have such a problem. And the problem is very interesting. Every 12vt rail with the exception of the one that's coming from the PC Express is shorted like every single one of them.

[00:00:15]So blown fuse blown fuse blown fuse. But every single phase as you can see here or the power stage or rail. Yes. So every single 12vt rail is shorted. How is this even possible? So given the fact that nor under normal circumstances, one driver MOSFET blows up, that one driver MOSFET is powered by one particular U rail uh one of the three and then you just replace the fuse, replace the driver MOSFET. Most likely almost 90% of the time the PMIC at the front or the back goes goes along with it and that's it. But in this case, all three are shorted. So, do we have like every single driver MustFet shorted or what's going on in here? Um, another thing that I could think of is a capacitor. there's a capacitor somewhere that's short and that capacitor happens to link well I don't know that doesn't make a whole lot of sense either because even the capacitors would not be see if if if the capacitor let's say one single capacitor were to filter any of these 12 volts that would mean that the capacitor would have to be connected to each one of these but that's not the case either because every single power uh every single 12vt rail is filtered separately so what do we I have no idea to be honest. Um, I'll inject some voltage and see if maybe something shows up on the thermal camera. I'm not seeing anything here. So, let's see. Yeah, nothing's going to be here or there or anywhere for that matter. Yeah, nothing here.

[00:01:49]Nothing there. So, we're going to have to increase going to have to increase the uh voltage most likely unless unless I start to see that this capacitor. You see this capacitor? See how how um let me turn off the lights here so it's a little bit easier to see. You see how all these caps are reflecting at the top? That's normal. But this guy, it's also reflecting on its side if I'm not mistaken. No, it's very difficult to get rid of the reflections. See it? Yeah, I think it was all just reflections. Never mind. Okay. So, I'm going to continue on looking. Um, I'm going to go ahead and increase the amperage or the Yeah. So, I'm going to give this thing a lot of amps. See if anything warms up this time. Yes. You see this? Yes.

[00:02:39]So, two capacitors. And what else does that is that the shunt resistor there? I think so. Yeah. And the shunt resistor.

[00:02:47]So, two caps. But that's messed up. Why would two caps are there any more caps?

[00:02:51]One more cap here. Holy cow. Can you guys believe this? So, two caps and a cap here. How does that make any sense?

[00:03:00]I have no idea. Let's go ahead and rip those P uh caps off the board. Okay, there's one. That's That's a lot of caps. As a matter of fact, it's rare for these caps to fail in such a way, at least uh based on my experience. But having all three, let's see if the shorts gone. Still short. Yeah, this one's gone. So, two are still short.

[00:03:23]These two are gone. Okay. So, that means a whole bunch of capacitors went bad.

[00:03:28]So, we've solved the short on this one.

[00:03:31]So, there's no more short there. But we're still getting short on the other two, which means more capacitors. Yeah.

[00:03:38]Okay. See right there? This this this guy. This guy. Yeah. So, the second one here. Here we go. Okay. So, short is gone. Yep. Short is gone.

[00:03:48]One more. Holy cow. This is incredible.

[00:03:51]I guess this thing is just aging. I don't know. This is messed up. I've never seen that many failures. Yeah, one more cap. One last cap here. That is That is balls, man. That's a lot of dead caps. Okay, so insert. Yep, short's gone. That's it. We fixed it. Let me see. Double check with the thingy here.

[00:04:10]Yep, it's gone. Okay, so what I want to do now, um, let's power on. See if it works. if it makes any sense to continue cuz it will work without these capacitors obviously. But um what I want to make sure is that there are no other issues such as blow and dryer musfets potentially most likely uh given the fact that the EVGA is constantly having power failure uh monitoring chip fail um failing uh when there's a short 12 we might end up having a problem with the power monitoring chip. So, let's see what the board is doing when it's powered on. And it's not doing anything.

[00:04:46]Wait a minute. It's doing something. You see, you guys see this? Oh, yeah. Yeah.

[00:04:50]Yeah. Yeah. Yeah. We got no fuses. Okay.

[00:04:52]So, let me go ahead and flip uh the solder the fuses and then uh we'll cuz I can't power obviously cuz fuses are blown. All right. Fuses replaced. The board is almost cooled. A little bit warm. Let's see what happens when we flip the switch. Oo, we got them shorts, ladies and gents. And we have another capacitor that's failed in here. So, let's get rid of that gently. There we go. And let's see. That's probably what happened. Okay. Wait a minute. Is that it? I'm going to I'm going to have to cool this thing down just a little bit more to see things a little bit more clear cuz it I thought I saw the driver Mustad that was failing, but after removing that one short cap, it looks like things went away. I'm starting to think I'm going to have to replace every 12vt capacitor on here at the very least the electrolytic ones. So, okay. So, that looks okayish. Yeah, look at that.

[00:05:49]Looks good. The PMIC 28°. Yeah, looks good on the power side. 2 and 12 amps more or less. Okay, let's see if this thing boots. I want to make sure that I can do some kind of basic testing before I spend enough time to uh replace all of the capacitors. That's not going to be easy or quick or anything in between.

[00:06:14]It's going to take a little bit of time.

[00:06:15]Thankfully, I have my uh new sucker delivered um recently. The Hakaco, the best sucker from Japan, ended up snagging one on eBay. gave the uh gave the customer or gave the seller a good offer that he couldn't refuse, which was about uh 95 to 98% of its original cost that he wanted.

[00:06:40]So, it was definitely an offer he couldn't refuse. So, who cares? It is what it is. A good tool will always cost good amount of money just like a good professional service will cost good amount of money. You know, though not always. There are plenty of crooks out there that are taking your money and giving you nothing. How did that happen?

[00:07:02]Like every single 12vt capacitor went well almost went kaboom. That is definitely not cool. How do they get dried up so badly? I have no idea, but I don't care at this point. So, I am uh currently monitoring the rest of them.

[00:07:17]Make sure that they're not going bad or anything like that. They all look okayish.

[00:07:22]So, and I think the card is posting.

[00:07:25]Yep, there we go. So, the card's posting 30° on the PMIC on the front. I think that's okay. And we're going to wait for the drivers to install. I want to make sure that all the power sensors uh reporting correctly. Then I will also look at the uh valley or something like that to make sure that I will not see anything that resembles things like that. That's a another uh EVGA 3090, the hybrid version that came from that same customer. We have the green squares and we have all sorts of these things and uh there and when I ran the valley or the heaven, I think yeah, whenever in the heaven it was all like rainbow flashes.

[00:08:12]So yeah, that's that's a dead core. All right, let's continue on waiting for this one. All right, here we go. First order of business is GPU to make sure that we can monitor any kind of overheating that could happen during testing. We do not want to overheat this thing. And you can already see that the temperatures are not in our favor because the cooler does not contact the core efficiently. That's okay. We're not going to be running firmware anytime soon. Okay. I want to see clock go up.

[00:08:40]We are Yeah, we always screwed up. You can see here we got 200 TDP. So, we're going to replace the PMIC's on the back and the front. But right now, I just want to see if we're getting any kind of weird artifacts and whatnot. Uh, I don't see any flashing. Okay, so that's good.

[00:08:55]Another test I can run quickly would be this guy here. Again, we're watching temperatures very closely. I want to see like two, three iterations. Yeah, that's probably enough. So, I think we're good.

[00:09:06]Uh, we are reporting gibberish on pin number two. It doesn't really matter where you get these reports, um, false reports from. And uh it's none of that's really matter that much anyway. Yeah, 400 watt on pin number two. Almost 400 watt on pin number one. Yeah, none of that makes any sense. All right, so at the very least, we know we have a working we have a working core, but we got a problem with the power management.

[00:09:31]So, that's great news. That means this is a repable 3090, which is kind of funny because I've just fixed it and only now I'm saying that it is repairable. And that's basically how uh my diagnostic process can go. Sometimes I actually have to get this thing to boot to post and run some tests and see if I have any kind of issues before I even get to say, "Yeah, yeah, it's repairable, you know. Um it is what it is. Not going to cry about it." So, let's put some standoffs here and there. Uh so what I want to do next is like I said I'm going to go and replace all of these uh caps.

[00:10:12]16 volt 16 volt 16 volt 16 volt. All of them are the same. Yes. So this one so all of the gray ones are the pink ones are for memory. So we're not going to bother with those. These are not 12 volt caps. They are all low voltage like uh 4.7 or something. 6.3 volt. 4.7 uh microfarad or 470. 470. These are all 270. Okay. And we're going to be replacing both of the PMICs, front and back, to save ourselves some time and a little bit of sanity. So, or we can replace the front one, see if it helps.

[00:10:46]And if it doesn't, uh, then we're going to take the front one and put it on the back. That way, we save at least one of the chips. So, we'll see. Who cares? Um, let's, uh, for now, I probably want to snap off the rest of these caps like so.

[00:11:02]and and just replace them all. That way, it'll be easier to desolder the legs.

[00:11:08]Although, not necessarily. I'm probably Yeah, it's probably not going to be easier, but I don't know. Just going to do it anyway. Who cares? I've never done it this way before, so I'm going to I'm going to try it and see if doing it this way actually makes it easier or not.

[00:11:22]Usually, I just leave them the way they are. So, and then so that the gravity when I heat the board just kind of makes them drop down on their own, you know, cuz when I desolder them, I just keep the board like this and then as I work my way across the board, they just fall down. And yeah, that was a bad idea. So, don't do that. It's just uh extra problems to deal with these pins that are not falling through. Yeah, just annoying. It is what it is, though. You learn something hopefully. So, I have to melt them. And then it seems like they're falling through, but they don't.

[00:11:59]Then I take a small soldering iron, push it through, and you would think that they fall through, but they don't. And then I take the pick, and then I run the pick all the way through. And you would think they would fallen through, but they don't. In this case, in this case, they actually did. So, who cares? And then you got all that low melt that you got to pick up that exit the other way around. Maybe even wipe it out. Yeah, probably be a safest thing to do. Just wipe it out. And then you still have to suck it all out. So, that's just another uh problem or a bonus. Where's my hako?

[00:12:31]Hako. So, while the hakaco is heating up, let's do these. This is probably the only place where I actually use low melt. Otherwise, I don't use low melt for anything. Sometimes I use this for a a power connector. Not all the time, though. Only when I'm in a super hurry.

[00:12:48]Uh, in this case, I'm not super hurry, but there's a lot of these uh capacitors. So, in order to save a little bit of time, I'm kind of cheating. So, yeah, I consider low melt a cheat. Now, with a dental pick, make sure we push it all the way through, which apparently does not give you any kind of confirmation that the you actually push that pin all the way out cuz it could still get stuck on the side as we will see in just a minute. And let's see. Yep. There's like half of them are still sitting here. So that's why this method was a waste of time. It' be more efficient if I left the capacitors intact. But breaking their legs from the beginning was the e that was the fastest way to sort of diagnose or rather start the process, but definitely not getting it finished. So with that done, see if we can suck it all out with our new toy.

[00:13:46]Oh, this thing works just magic. So good. Wow, that's just beautiful. My old sucker would not do that. He would get clogged up nonstop. But this one seems to be holding up very well.

[00:13:58]I think they they when the Chinese try to copy it, they didn't really copy the tip. The tip that it's on the old one is just just a hole. Whereas in here, it's not just a hole. It's actually different diameters.

[00:14:11]So, The other one would get clogged all the time. Wow, that looks really good, I think. See if I can catch any solder on the Nope. Looks good. It looks absolutely perfect. All right, so now a whole bunch of caps and uh we should be done.

[00:14:34]All right, so the caps are replaced. Uh I didn't replace the uh PMIC, not yet. I think I want to go a slightly different route. So, power is still good. I want to go and see if I can narrow down which one of these chips is bad. A lot of times it's not that easy because you can be getting getting a good readings, but the chip is faulty nonetheless. So, we're going to start checking some pins.

[00:14:56]You probably cannot see it. I had some flux in here. I was about to remove this chip and I thought to myself, maybe we can just narrow it down. And some people might even learn something. All right.

[00:15:06]So, we're looking for 12 volt on pin one, two, uh, four, five, seven, eight, and then exact same thing on the back. So, 12 volt, 1199. So, that's going to be our baseline. 1195 or whatever. Close enough. And then on the third pin, we get 0.77. Okay. So, next uh three will be exactly the same or at least should be. So, 188. So, are we sagging in here?

[00:15:32]187. We're going to check with this one here. Nope. We're not sagging. We're just warming up. Okay. So, 1132. Oh, that's definitely different. So, is it 1132 on the side of the chip? Yes. Or is it the 1132 on the side of the resistor?

[00:15:45]It's 1132 on the side of the resistor.

[00:15:47]So, is it this resistor here? Do I get 12 volt or the 1186 or whatever? Same as I get here. So, 1185. Not sure why it's reading 1185. Let's switch the probes around. 11.92. One more time. 1186.

[00:16:02]We're a little bit dipping down here, but everything else, no, we're still dipping. We're dipping on the end of the resistor. Uh, and let's check one more time. 11.87.

[00:16:13]Why are we dipping so low? 3 amps. We're climbing up to 3 amps. So, maybe I give this thing more amps. Let's give this thing more amps. One more time. What are we reading on the coils? Coil is reading 119. 118. The other one is 11. Okay. So, 118. 119 is what you should be getting around here. Yeah. So, it's about the same. We're looking for more dramatic differences, I guess. 11 9 11 88 almost to 9. Whatever. And 0.26. I think this pin supposed to be Whoops. 0.26. Okay, never mind. So, 1 2 3 0.26 1 2 3 0.26.

[00:16:51]So, 12 or the exact same 11.8. Exact same 11.8. Oo, 11.2. Uhhuh. See, this one's a lot lower than this. See 11.9 or 11.8, but this one's like 11.2. Yes. So, that's a problem right here, definitely.

[00:17:07]And this pin is going to be very different. Oh, yeah. So, this pin, where does it go? This goes into here. This resistor here. So, we got good voltage here. And we got bad voltage after. It's not the resistor. It's this guy. All right. The chip replaced. Let's see if the readings are changed. So, we're going to start. I don't know where the chip came from. So, we're going to start from the very beginning. So 125.

[00:17:31]Wait, what? Where did that go? 1213.

[00:17:34]About the same. This one can be measured here. 0.25. So that's good. 12. 12. I think we go here. Yep. Same thing. What about on this end? 12. 12. And this one here, I believe. Yep. Same reading. And one last one. 12. And that this is where we had trouble before 12. So now it's 12. And the last one exactly the same where used to we used to have like a 0 point uh 1.6 volt or something like that. And now we're reading good voltage. So maybe we got lucky and uh the very first chip that we replaced will end up being the one that was faulty all along. So who knows? Let's boot into the Windows and uh see if anything changes. Now, if you're still not sure what my me what am I measuring, what for, what am I looking for, and how this process works. I have an earlier vers uh earlier video about the exact same card where I explain things in a little bit more detail. So, don't be shy. Go check it out. I don't know what it's called. Um so, maybe somebody knows, you can post it in the comment section below what the title of the video is to make it easier for the rest of you. All right, GPUZ. And we can probably run ME test as well. Okay, so what's our TDP? Oh, look. TDP is normal.

[00:18:55]12 volt. 12 volt. 12 volt. 6 watt. 8 watt 12 watt. Okay. Let's uh run this guy. See what See what it does. 30%, 99%. We are overheating, but everything is looking good. Close. Close. Close.

[00:19:08]Close. Close. We do not want to overheat. Okay, we're done here. This GPU is 100% fixed. All I have left to do is Well, it's 99% fixed. I still have this 1% chance that we may have to rebol the core. Who knows? So it might fail and start crashing. Uh I won't know until I run all the tests. So even after all that work, I am still in the diagnostic process of this thing, believe it or not, until I get it done completely. So until I get it all fixed.

[00:19:38]And uh I don't know, kind of makes me wonder when people pay me for diagnostic, do they pay me for diagnostic?

[00:19:46]cuz cuz the way I do the way I I deal with this is that diagnostic and the repair attempt it's one cost. So let's let's say that right now everything appears to be working and and I hope it is and then I start running a test and I start to see like flashes and geometric distortions and games cuz I haven't ran any games other than the valley or heaven or whatever. So I still don't know how other games will perform. And imagine if val imagine let's say uh time spy will give us some kind of geometric distortions or whatever that would render the core dead and after all that work it would still be the same diagnostic or repair attempt cost. So I don't know I think it works. I think it's I think it's fair to uh customers because you know exact same thing would have uh is happening to the other 3090 from that same customer. I diagnosed it in like 5 10 minutes just looking at the artifacts and uh but in this case I still don't know with all that work that I put in. I still don't know whether the core is working or not. And if the core ends up not being working it's just a repair and diagnostic fee. Uh repair attempt diagnostic one fee and that's it. I think it's fair. Uh, but let me know what you think about it. I'd like to hear you your uh thoughts on this.

[00:21:05]And uh that will probably be it for this repair. I'm going to go ahead and reassem uh reassemble everything. Put the fresh pads and uh paste everywhere where it needs to go. And I hope I earned your like and maybe subscription if you're into this kind of stuff. And uh please don't forget to leave a comment to help with the algorithm. And have a blessed day. Goodbye.