Might be possible to replace it, damage doesnt look beyond repairing, will need to replace both the mosfet and the A2724 fet its linked to.

Be glad you were there when it decided to let out the magic smoke, could have been much worse !

Boy I hate that smell. I distinctly remember that burned Socket A CPUs had a sort of weird fruity smell in the mix.

It doesn't look bad from a distance but the outer packaging material had actually carbonized and was brittle to the touch. I chipped away at it but could not remove the inner transistor mass from the board no matter how much flux, hot air, or heated tip from my iron that I used. The bottom of the chip had fused to the copper ground plane beneath it. I finally ended up using enough force that I tore the entire copper surface off of the board, revealing the fiberglass layer beneath it.

Since the board did initially turn on after the smoke was let out, I guessed that after the chip had burned up that the system might still function without it, but I wanted to remove the chip before trying to power it on again. But at this point I have no idea what kind of damage I've done trying to remove it, and too I'm afraid to plug it in again to risk damaging other components.

It's dead. Oh, it's definitely very, very dead. You can see from this photo that the fiberglass beneath the ground plane had also turned black as a result of the chip burning up.

It probably welded itself to the ground plane, now that I think about it. Even after tearing it off and away from board, it would still not let go. At this point it should have slid right off.

I've never seen a motherboard with a Molex connector on it. According to the manual this needs to be connected if you are running two video cards.

Errius wrote on 2022-03-26, 06:39:

I've never seen a motherboard with a Molex connector on it. According to the manual this needs to be connected if you are running two video cards.

A friend of mine with a board like that had misunderstood its purpose and used it to power a fan using a molex->3 pin adapter with a pass through connector.

Errius wrote on 2022-03-26, 06:39:

I've never seen a motherboard with a Molex connector on it. According to the manual this needs to be connected if you are running two video cards.

I always felt it was pretty common for boards from that time to have that connector for additional power for PCIe.

Oetker wrote on 2022-03-26, 06:48:

A friend of mine with a board like that had misunderstood its purpose and used it to power a fan using a molex->3 pin adapter with a pass through connector.

Oh, cool, so it works both ways! 😄

I would be interested to know what causes such an issue. The power draw on the CPU side can be significant, but you would hope that the engineers include failsafe components to reduce or remove the risk of overheating.

I had a similar experience with a HIS 7950 where on one of the VRM rails a component exploded spectacularly during a hashcat operation. The safety margins are probably kept thin for "commercial" reasons...

flupke11 wrote on 2022-03-26, 07:08:

I would be interested to know what causes such an issue. The power draw on the CPU side can be significant, but you would hope that the engineers include failsafe components to reduce or remove the risk of overheating.

I had a similar experience with a HIS 7950 where on one of the VRM rails a component exploded spectacularly during a hashcat operation. The safety margins are probably kept thin for "commercial" reasons...

Its a fairly old board that was never top tier even when new so it like was built to a budget, as such it likely wouldn't have any crazy protection circuitry on the VRM components, I would expect very basic protection. Being a P35 it wont be hard to get a cheap replacement, if it was a X48 it would be a more painful loss.

I bought this card for testing SLI cards, but I only had one card inserted when it wasn't posting at this time. I even tried removing the video card and turning it on with just the motherboard, CPU, and memory, and it still wouldn't come on.

The original issue was that I had damaged a video card in this board by inserting it with an open memory slot. One of the latches on the slot met perfectly with some capacitors on the back of the card and, when it wasn't inserting all the way, I applied some force and *plink*.

I tested the card as working before the repair in this board (it's just a bypass cap), put the capacitor back on the card, popped it back in the motherboard, then it wouldn't turn on. I was confident that it wasn't a result of my repair, especially when it still would not turn on with another video card, or any inserted at all. I also had no problem testing this board with a couple of cards yesterday, so it should have been working.

The piece that fell off was a 1206 47uF capacitor. I didn't have that value handy so I reattached the original. It measured about 42uF, which is within the margin of +/- 10% and seemed fine. It was not shorted before or after reattachment. Does anyone know what voltage a capacitor like this should have for a replacement?

Doornkaat wrote:

I always felt it was pretty common for boards from that time to have that connector for additional power for PCIe.

It never was, only somewhat high-end boards have it. Which is a shame, because additional connectors can also improve voltage drop on PCIe slot with one video card.

well it was somewhat common on high-end boards so to say ^^
and it still is, only it has since been replaced with a PCIe power connector.

I have seen some issues with some "Epitaxial" mosfets that for some reason, after working perfectly for a long time even during extended periods of time, decide one day "Allright, I've had enough with this, this is the end for me" and boom, they bring down the rest of the equipment with them. I think its an issue that takes time to develop though, because, i've only seen it happen on equipment that is stressed. For example, a customer would come to my house/shop and tell me they've upgraded the processor to a faster one lets say, 6 months ago, but this is a computer used to not having to deal with a beefy processor in the first place (The case I mention is for example, swapping a core i5 to an i7) and when troubleshooting, sure enough, the power rail for the processor is shorted. Usually, just changing the mosfet solves the issue because the PSU did detect an overcurrent and shut down, but I'm afraid, you werent lucky enough and the juice kept coming. Sorry to see that board gone.

flupke11 wrote on 2022-03-26, 07:08:

I had a similar experience with a HIS 7950 where on one of the VRM rails a component exploded spectacularly during a hashcat operation. The safety margins are probably kept thin for "commercial" reasons...

You got to pay more for the "OC" boards with more robust components. It reminds me of the Gigabyte X58A boards, the UD3R, UD5, UD7 and UD9. There are few UD3R boards around still in working condition because they tend to die quickly. The others are tanks.

I'd guess that the dead FET was on the high side of the power supply and failed open. The output voltage would then be too high, so the controller would have the low side FETs on all the time, so a lot of current would flow through the failed FET, then to ground through all the low side FETs. What I think is the input inductor also looks like it got a bit toasty. If the controller managed to keep the voltage from rising too far then there's a small chance the board might still be ok, if messy. The via by the Gate pad (pin 4) looks to be intact, and if you're lucky there weren't any signal traces routed under the FET. Check for any connections between the nearby 6609 (FET driver chip) pin 1 (UGATE) and pin 5 (LGATE) to the failed FET pin 4 pad and also pin 4 on nearby A2724. If the gate connections are intact then you might get away with dropping in a replacement, although the centre pad connection will be missing, so it might need a heatsink of some sort. Also check for any shorts between drain and source pins (pin 8 and pin 1) on the other FETs.

From a photo of another gigabyte board then I think the failed FET was an A2726 ( http://rom.by/files/UPA2726.pdf ). Maybe something like https://www.mouser.co.uk/datasheet/2/115/DIOD … 8_1-2543300.pdf or https://www.mouser.co.uk/datasheet/2/115/DIOD … 8_1-2543300.pdf would work as a replacement.

I'm not keen on trying to power this board on after having butchered a hole through it. Even if I could get it working, I would be afraid of it being a fire hazard or risk that it might fail catastrophically in the future and take some other component with it.

If I could have gotten the MOSFET off, I would save tried. But not anymore.

This is actually the second time I've had a computer catch fire. The first instance of pyrotechnics was entirely my fault, unlike this suicidal MOSFET. I had inserted a memory stick into a machine while it was on its side and hadn't seated it fully. When I powered on the machine it shorted some pins and a pillar of smoke to erupt from the case. The memory was charred and the slot was melted, but surprisingly the motherboard still worked with the other memory slots afterwards. I did eventually die some months later, possibly as a result of this trauma.

Kahenraz wrote on 2022-03-26, 20:20:

I did eventually die some months later, possibly as a result of this trauma.

I see ghosts!

I now have further evidence (sample size of 2) that these kinds of pyrotechnic events can lead to the component welding itself to the board. This makes repair extremely difficult, with a high chance of the board becoming damaged during component removal. The danger here is that due to the welding process, it's no longer possible to remove the component with heat alone. And the necessary mechanical stress can easily tear the board apart, especially since the heat from the event may have already caused the layers to separate.

Re: eBay steal turned out to be a charred wreck

I don't have any more boards to experiment with, but I will instead approach these kinds of repairs very differently in the future. It may be safer to remove only parts of the component or sever traces and instead bodge a replacement nearby instead of risking catastrophic damage while trying to remove the component. This is much more difficult for some repairs, such at this motherboard BGA MOSFET, as the component pins are tiny and difficult to work with, and there is no physical clearance or position for adjustment as there is meant to be a heatsink mounted directly on top of them.

For this specific repair, it might have been better to chip away the carbonized material and sand away as much of the welded material as possible, since it was pretty obvious early on that heat was no having any effect. Of course, I wasn't certain of this and decided to press on and test exactly what the result would be; which ended up being a hole in the board. At least my second attempt was much better! 😀

kahenraz, I think you forget the biggest annoyance with components that have overheated and charred the board is that internal layers can now be connected with some resistance due to the carbonized remains of the board.

That's a good warning. I don't think there is any good way to test for this without ripping up the top layer and digging underneath. I will be testing for shorts between the voltage rails and ground plane, but at some point I'll have to plug it in and see what happens. I have a certain unloved Dell motherboard I use for these kinds of risky tests.