[Solved] RX 460 Blown Fuse

I could buy it and try it. Does it have to be A4N? Or can it be anything "NCP81161" like A4F, A4L etc.? I guess it's just a marking on the same thing.

I removed the card today to check if these are indeed A4Ns (they are). I didn't do that previously as I thought it could lead to some problems and it did. After plugging the card back, the PC wouldn't start with it (fans spin, but card wasn't detected). Few starts later still the same thing. Turning AC on/off didn't change anything. Only after clearing the CMOS, the MOBO made some kind of self check (like after every CMOS clearing) restarted itself and booted successfully displaying picture. So I guess these A4N (NCP81161) chips are a good guess. Unless it's the capacitor ;) To check the cap I guess I should remove two of them (for each phase) and measure them to see if they have the same values (because they should?).

BTW I'm trying to understand something here.
I mentioned NCP81022 (4 + 1 phase) and it's one chip in the front. You said NCP81161 and they are there too (on the other side where you marked them).
In the front of the card I can see 4 + 1 phases (high and low side mosfets). On the other side these 4 phases have low side mosfets only (marking K0393).
Do I guess correctly that they are connected with the high side mosfets in the front, so the current given by high side mosfet is split between two low side mosfets?
View attachment 242334
It doesn't have to be an A4N. The N is a date code to indicate when it was manufactured. Any A4 should work, assuming it's actually an NCP81161. To check the bootstrap caps, you would remove the one on the suspect phase and check its resistance and capacitance. You should have near infinite resistance across it, and capacitance should be 0.027uF. If you have different capacitance, remove and test one of the other caps from a different phase. If the capacitance matches, they're probably OK. If you don't have pretty high resistance, it's bad and needs replacing.

It's hard to say for sure how all those mosfets are wired up without the card in hand, but it's pretty common to have two low-side mosfets and one high side. It totally wouldn't surprise me if that's how this card is wired up. You could check by testing to see if you have continuity between the pins on the pairs of low side mosfets.
 
So I'm back to tinkering with this RX 460 :D

It doesn't have to be an A4N. The N is a date code to indicate when it was manufactured. Any A4 should work, assuming it's actually an NCP81161. To check the bootstrap caps, you would remove the one on the suspect phase and check its resistance and capacitance. You should have near infinite resistance across it, and capacitance should be 0.027uF. If you have different capacitance, remove and test one of the other caps from a different phase. If the capacitance matches, they're probably OK. If you don't have pretty high resistance, it's bad and needs replacing.

It's hard to say for sure how all those mosfets are wired up without the card in hand, but it's pretty common to have two low-side mosfets and one high side. It totally wouldn't surprise me if that's how this card is wired up. You could check by testing to see if you have continuity between the pins on the pairs of low side mosfets.
It's exactly how this card is wired up. There's one high-side and two low-side mosfets for each phase.
Bootstrap capacitor seems to be ok. I can't measure capacitance but I checked the resistance and it's pretty much the same as any other bootstrap cap on this card.

I swapped NCP81161 on the faulty phase and it seems to have helped. So far 4 days of heavy use and nothing is shorted. I will report if something will happen with the card.
 
Well, it died again. It worked fine for over a week, and today suddenly it just died on the desktop. So obviously it's not only a mosfet, not the NCP81161 either.
 
Well, it died again. It worked fine for over a week, and today suddenly it just died on the desktop. So obviously it's not only a mosfet, not the NCP81161 either.
Did you try replacing the bootstrap capacitor?
 
I didn't, but I can replace it and see if that will help.
That's what I'd try next. Apparently, there's a pretty common failure mode for MLCCs where they can crack internally such that they become either an open circuit or a short, but only while they're running, and this is made quite a bit worse with heat.

The bootstrap caps on a lot of cards are often mounted in such a way as to be subjected to basically all of the potential causes of this.
 
I replaced mosfet and fuse again and I also replaced the bootstrap capacitor. As always, card is working now, but the question is for how long? Additionally, I checked the voltage from bootstrap pin to ground this time as mentioned earlier in this discussion. I found that when card is on idle it's 0.6v, under load ~0.8v (depends on the load).
 
What do you get across the bootstrap caps on the other phases?
There's definitely something wrong with the bootstrap circuit on this phase when you compare these voltages to the other phases. On the phase below the faulty one, it's ~12.9v on idle and ~13v under load. That's a massive difference, but what is causing this? Also one interesting thing is that there's voltage on that pin even when computer is cut from power but it's not the case with the faulty phase (faulty cap?). I swapped bootstrap capacitor so that's weird.
 
  • Like
Reactions: Kzoak
like this
So, you're measuring between ground and the bootstrap pin on the phase controller, and on the good phases, you have 13V, and on the bad phase you have 0.8?
 
So, you're measuring between ground and the bootstrap pin on the phase controller, and on the good phases, you have 13V, and on the bad phase you have 0.8?
Exactly. To be precise I'm not measuring from the leg of the NCP81161 because it's too small. I measure it from the leg of the bootstrap capacitor that is directly connected to that pin.
 
Does this card have the 0 ohm bootstrap resistors between the actual pin and the capacitor?

Also, what do you get if you put your meter in diode mode, and measure between the bootstrap pin and the 12V input to the card?
 
I get open circuit between the bootstrap pin and the 12V input - both on good phase and the bad phase There are 0 hm bootstrap resistors between the pin and the capacitor - my multimeter is showing around 1 ohm there on the good and bad phase.
 
Well, I didn’t thinked of that. Reversing the probes helped :p There’s 1371 on the faulty phase and 257 on the good phase.
 
Well, I didn’t thinked of that. Reversing the probes helped :p There’s 1371 on the faulty phase and 257 on the good phase.
OK, let's try something else. Put your meter in resistance mode, and measure resistance from the 12V input (red probe) to the bootstrap pin (black probe). This is called the forward direction. Then, reverse the probes and measure again. This is called the reverse direction.

The expect resulted is a few K in the forward direction and millions or infinity in the reverse direction. Try this on all four phases, and see what you get.
 
  • Like
Reactions: Kzoak
like this
OK, I did the measurements and here's what I've found:

Forward direction
Good phases - ~14.5K
Bad phase - 54.1K

Reverse direction
Good phases - ~2.89M
Bad phase - 2.95M
 
It might be possible that the previous bootstrap capacitor was indeed faulty and did something to the changed NCP81161 on the faulty phase. Therefore I'm getting those weird values. Just guessing. I can swap the NCP81161 again, but I don't know whether it's even worth trying that.
 
It might be possible that the previous bootstrap capacitor was indeed faulty and did something to the changed NCP81161 on the faulty phase. Therefore I'm getting those weird values. Just guessing. I can swap the NCP81161 again, but I don't know whether it's even worth trying that.
That's probably what I'd try next. I'd at least remove it, and check the resistance on the VCC and bootstrap pins. You should have one diode drop's worth of resistance in the forward direction, and infinity in the opposite. If you have a known-good NCP81161, you could also swap that on there.
 
  • Like
Reactions: Kzoak
like this
That's probably what I'd try next. I'd at least remove it, and check the resistance on the VCC and bootstrap pins. You should have one diode drop's worth of resistance in the forward direction, and infinity in the opposite. If you have a known-good NCP81161, you could also swap that on there.
Well, that was exciting. So I swapped NCP81161 and the bootstrap cap and after that forward direction resistance was 13.5K so close enough to that 14K of a good phase. Then I connected my multimeter probes to measure bootstrap pin to ground voltage and I turned on the computer. Multimeter showed 11.97V and then inductor on the faulty phase just burst into flames. I don’t know why though. Maybe one of the pins of NCP81161 didn’t solder properly? It’s really hard to solder this SMD package. I will swap this inductor and try to resolder that NCP81161.
61ABB0BA-6A30-40AB-9050-EAE131DBD079.jpeg
 
So it was just a poorly soldered pin that caused that. I did manage to run it after fixing that and replacing the inductor and also mosfet and voltages and resistances on the faulty phase were the same as the other phases. But it died again anyway, even faster within few seconds of load. This card has a really weird issue.
 
  • Like
Reactions: Kzoak
like this
I think that it might be just a PCB that goes bad after waranty time due to insufficient cooling of VRM. I literally tried swapping everything on this phase and result is always the same. I guess this card just can’t be fixed.
 
I think that it might be just a PCB that goes bad after waranty time due to insufficient cooling of VRM. I literally tried swapping everything on this phase and result is always the same. I guess this card just can’t be fixed.
What seems to be confirming that is that on the faulty phase there is a higher (about 1K higher) Gate to Ground resistance on my card. From what I can see this resistance is mostly depending on a resistor between high side gate driver output and ground (R510 on the faulty phase). Resistor itself is good and has the same resistance as the other's. So that additional resistance seems to be caused by a PCB itself. NCP81161 is good and mosfet too. I might be wrong though.
 
What seems to be confirming that is that on the faulty phase there is a higher (about 1K higher) Gate to Ground resistance on my card. From what I can see this resistance is mostly depending on a resistor between high side gate driver output and ground (R510 on the faulty phase). Resistor itself is good and has the same resistance as the other's. So that additional resistance seems to be caused by a PCB itself. NCP81161 is good and mosfet too. I might be wrong though.
So, one side of R510 is supposed to be connected to ground, but it has a non-zero resistance?
 
So, one side of R510 is supposed to be connected to ground, but it has a non-zero resistance?
Not exactly, resistance on the ground leg is near zero. But something happens when measuring through this resistor and resistance is higher than the corresponding resistors on the other phases. But resistor itself is good.
 
How "near zero" are we talking about? Any idea what that leg is actually connected to, if not actually ground?
 
How "near zero" are we talking about? Any idea what that leg is actually connected to, if not actually ground?
It’s 1.8 ohms. There might be something in between actually when I think about it now, don’t know what though.
 
Is the 1.8 ohm side perhaps connected to the GPU core? Check for continuity to the vcore VRM's output chokes.

But, to be clear, you've got four phases that have this resistor, with 1.8 ohms to ground on one side of it, and a few K on the other side, but on the suspect phase, you've got 1K higher than the other three?
 
Is the 1.8 ohm side perhaps connected to the GPU core? Check for continuity to the vcore VRM's output chokes.

But, to be clear, you've got four phases that have this resistor, with 1.8 ohms to ground on one side of it, and a few K on the other side, but on the suspect phase, you've got 1K higher than the other three?
It is actually, to every output choke. Same for the other resistors. There’s continuity there. And yes exactly. On the other phases there is 7.7K -7.8K but on the suspect phase there is 8.7K.
 
It is actually, to every output choke. Same for the other resistors. There’s continuity there. And yes exactly. On the other phases there is 7.7K -7.8K but on the suspect phase there is 8.7K.
On the 8.7K end of that resistor, you have continuity to the high side gate pin on the MOSFET and the high side gate drive on the 81161?
 
On the 8.7K end of that resistor, you have continuity to the high side gate pin on the MOSFET and the high side gate drive on the 81161?
There is continuity to the high side gate pin on the mosfet and 1.6 ohms to the high side gate drive.
 
Yes, it’s the same on every phase.
I think what I'd do next, then, is to remove the 8K resistor from the suspect phase and one of the good phases and measure them both. If the resistance isn't the same, replace the one from the suspect phase with a new one that matches the good phase.
 
I think what I'd do next, then, is to remove the 8K resistor from the suspect phase and one of the good phases and measure them both. If the resistance isn't the same, replace the one from the suspect phase with a new one that matches the good phase.
I did that and those resistances are the same, 10K for each.
 
I did that and those resistances are the same, 10K for each.
From what I found, 81161 is decreasing resistance of this resistor (without it resistance is 10K, just like after desoldering this resistor from the board). But on this phase even with a 81161 from the other phase it still has that 8.7K resistance.
 
Back
Top