Old hardware, handled on the desk after all this years. All it would take is one ball under Piix4 to crack forcing the chip to latch up and there you go. Thankfully its a cheap and available IC.

Thank you for your reply, I am having second thoughts now, if I just supply the 3.3v rail on its own, with no ATX power even if I set the bench PSU at 2 amps, the 3.3v rail is only consuming 1.11 watts so around 370 Ma and the south bridge does not get hot, it warms up to around 32 degrees c but not 70 or 80 when putting 1.5 amps into it when the ATX is powered up as well when it was pulling nearly 5 watts or around 1500Ma. So I am now wondering why that is the case.

I would not be able to fit a BGA as I don't have the equipment yet to do it and have never done it before but there is a first time for everything I guess. But a useful tip, it might be worth applying pressure to it, or re-flowing, but curious why its not showing any symptoms with only the 3.3v rail?

Its really REALLY bad for chips to run without power while still connected thru IO to power source. Failure mode is called latch up https://en.wikipedia.org/wiki/Latch-up when IO sees higher voltage than VCC.

If your chip cooked at 80C then it might be game over for it. Since it cooked in the first place then not all VCC break, but something else like Vref or one of the grounds or just on of VCC balls. It it broke in another way entirely.

limsolo wrote on 2026-05-17, 16:49:

Hi All, […]
Show full quote

Hi All,

OK so dead FW82371 southbridge, injecting 3.3 volts at 1500Ma into the 2SK2941 3.3v MOFET (corner near DIMM slots) source under thermal camera it (southbridge) hits 80 degrees after around 10 seconds. Not sure how or why this happened. Here is the story.

Board working fine, swapped out the ICS 9150 for a 9250 board working fine, still could not O/C my dual PIII Slot 1 850's though, all I wanted was 112MHz bus which it would not do with the 9150 (bug) so still would not do it with the 9250, so was looking at a way to up the CPU voltage around 0.1 volts and see if that would do it, anyway board is on the bench and I am reading a datasheet on my lab PC, Windows 2000 blue screens and then the board is dead, no life at all, I check all powers and ground at ATX connector all good, so anyway I find primary CPU is dead/cold but secondary CPU is getting warm, so I check and the MOFETS for second CPU giving correct voltage but primary nothing zero, so I swap CPU's, cold CPU that was in primary in second slot gets warm. So I check HIP 6019 and nothing conclusive so I fit a new one - no change still not power to primary CPU. I should also mention CPU fans do not run.

So PC Analyzer card in ISA and PCI no 3.3v rail so find the 3.3v MOSFET as mentioned above, source nothing coming out but 5v going in the drain, and gate is getting around 2.6v not high enough but not dead either, so lots and lots of checking, scope the 9250 and that is giving PCI clocks 33MHz, 24 and 48MHz clocks (I thought maybe the new 9250 was shorting and pulling down the 3.3v rail) but it looks like its alive from the scope.

So I then start to inject 3.3v into the source leg of the 3.3v MOSFET with a bench PSU, first 200Ma, then 300Ma at around 600 Ma the CPU fans start to run, at all the time I am looking at the board with a thermal camera, nothing odd, room is 25C and it this point the hot spot is the Piezo speaker at around 30C (???) so then I get up to around 1200Ma, still nothing hot and at this point I get a two tone error from the Piezo which repeats low > high > low > high and cycles like that (and I cannot find any info on what this error is) so anyway just got to 1500Ma (still set at 3.3v) DMM meantime the source pin is showing 2.9v so 0.4 volt drop because of the load but at this point the alarm went off on the thermal camera. at 1500Ma and 3.3 set 2.9 actual it was pulling nearly 5 watts.

So shorted southbridge pulling the 3.3v rail down, I think I will just for the sake of it remove the southbridge just to make sure the short is gone, logically it should be. I should also point out at this stage, all my soldering work is fine, checked under a 4k camera/scope and DMM all over the place before hitting the board with voltage injecting showed no shorts to ground or low resistance anywhere so all basics covered, also in this state doing injection, no PCI or ISA cards, no drives, no ram, no CPU's or fans fitted, just the bare bones board plugged into an ATX supply (on) and the bench PSU supplying the 3.3v rail manually.

So question is had anybody seen a P2B-D or (DS) or even a P2B or another make/model of 440BX board die all of the sudden because of a fried southbridge?

Thank you

Yes, the Asus P2B can burn out the SB, as is often the case with the P5A (ali), and even a BX440 can be damaged if it isn't cooled, while the LX440 can run bare, and some boards died due to underestimating the BX440's heat dissipation!
However, the southbridge was NEVER cooled, and therefore can break much more easily than the northbridge!

But how is the southbridge normally powered in the P2B-D?
Does it heat up immediately, or does the motherboard work for the first few minutes?

Thanks guys, for your feedback much appreciated and I will definitely read that article about latch up, diagnosis is going to change or stop for the time being on this for some reason (this is after I had decided to check all the pins of HIP6019BCB) to ground with a ohm meter the board sprang into life, not with CPU 1 installed but with CPU 2 installed (will not post with anything in slot 1 right now), I saw the 3.3v rail come up first (well heard the CPU fan start for the first time in a long time), and the two tone error, saw CPU 1 voltage was low at 1.3v and 3.3v rail low at 2.8v, so put the CPU in slot 2, and now it posts on CPU 2 no error tone and can get in the BIOS screen.

I know why 3.3v is at 2.8v, I lost the 114 (110k) resistor and subbed a 100k instead this resistor is part of a voltage divider for the sensing of the 3.3v rail (PIN 11 HIP6019) so I think that is what is upsetting things at the moment. So going to fit the proper 110k resistor and that will fix the 3.3v rail and I recon also the HIP6019 will be happy enough to start giving 1.65v to CPU 1. I wanted to see if the SS on pin 12 was being held low on the board because of a fault or held low by the chip, so wanted to pull the cap of PIN 12 and see, but blew the 114 (110K) resistor off the pad on PIN 11 across the workbench at the same time 😂

So I was probing HIP6019 under a binocular 4k microscope, I know I can solder to a good standard and I was probing each leg to ground - I know for a fact that changing the chip fixed nothing and I know my soldering is fine having checked it with a loupe and done a pin push more than one time. I know all the high and low and other buck converter mosfets are not cracked or overheated, I know they are soldered in place good.

So doing an ohms test on HIP6019BCB "woke" something up but not the chip itself, something connected to the chip. I wish I could a bit more scientific than that but I reckon as soon as I fit the resistor tonight and the 3.3v rail is fully up and it will post fine on both CPU's and there won't be anything to diagnose until I am wrong or it comes back in the future, bearing in mind, it was working great until was poking around the high and low CPU 1 MOSFETS.

Without being able to say anything conclusive, it would seem that probing the high and low mosfets on CPU 1 did "something to one or both of them" what and why I have no idea, and as it works now or at least partially, I think it will be hard to say why.

I will update in a day or two

Were you probing powered On system? in OHM setting? 😮 thats a big no no 😀 Probing mosfets in meter diode mode can turn them on/off. Maybe HIP6019BCB is bad and wasnt driving one of the mosfets at all leaving if floating somewhere in linear state, that would make whole thing short every cycle.

Haha no I know that - when I was trying to O/C I was measuring the source voltage (using Uni-T meter UT61E+ not the best as my Fluke is not working at the moment but not crap either) out of the MOSFETS for CPU 1 everything was fine board was running fine on the bench on a large silicone mat Windows 2K was up and I was looking at a datasheet it was like that for maybe a minute and I think (its a couple of months ago now) it blue screened and that was the last time until last night there was any life out of it. Last night board was off, unplugged from ATX or bench power and I was in ohms mode, same meter, going around all the pins on HIP6019BCB and HIP6004BCB) making comparison readings from one set of values to the other to see if I could spot anything very different - which I did to be honest, anyway I thought OK I am going to take voltage readings now (the reason I did this was I had never looked at the 6004 as that CPU was getting warm so I thought what is the point, but then I thought let me take some back to back readings same circuit CPU 1 vs CPU 2) anyway before I got a chance to take any readings as soon as I powered the board on the CPU fan was running so I measured the 3.3v rail at 2SK2941 source which is the MOSFET for the 3.3v rail and saw 2.8v before this had been 0v even with 5v at the drain and nothing at the gate (so this MOSFET was not being driven before), I didn't see any post codes on the analyzer card but knowing this board will usually run fine with a CPU in any slot and knowing CPU 2 has got 1.65v I moved the CPU over and the board posted.

Like I said the board was dead, I changed 6019BCB for another new one, still the same still dead. I know you can turn MOSFETS on and off using a meter in diode mode, not sure about other modes, but I think your explanation is the only one that makes any sense, it faulted when I used the DMM in volts mode, then un-faulted last night when I was in Ohms mode - unless is an intermittent fault which was be a big coincidence then there is no other logical conclusion other than what you say and that is what I was thinking anyway - the only other thing I can thing of is, and I don't recall doing this but at close to 60 my eyesight it not what it was, perhaps I shorted a MOSFET drain source or gate drain momentarily and didn't realize the the MOFET got stuck ?

Meter brand doesnt matter much, we arent calibrating precise instruments.
Oh so measurements when off. Definitely a coincidence.