A lot of these old motherboards are getting oxidised or even rusted pins in the cpu socket. I would try some electronic cleaning spray (the type that drys) in the socket pins and then with the
cpu in the socket open and close the lever a few times.
Also try the memory in the other slot and clean the edge connector with an eraser.
Caps is unlikely if it occasionally starts when cold, but as caps usually fail when cold, try heating them with a hair dryer before trying to turn the motherboard on.

Do You have other memory or cpu You can try?

edit. I see You mentioned an issue with the cpu and heatsink

Pics of the heatsink, pls.
Warm it and remove the lid using a fine metal thread. Requires a vice. You can try without warming first.

While the board is running, measure the voltage between the MOSFET output pin and ground. It should be close to the expected core voltage of the CPU.

If those are constant speed fan connectors, no PWM or voltage based control of the fan speed, then it's not good for the fan speed to be fluctuating. That would mean the +12V supply is fluctuating.

As suggested above, by probing the mosfets that feed Vcore, or the inductors, you can probably find what Vcore is actually being supplied. Is it correct and is it stable?
Try probing the +12V supply also - verify if it's stable or fluctuating (as the fan implies it is).

Is this a Prescott CPU? Those CPUs were especially hard on VRMs of many motherboards because they are so power hungry. If it's been overclocked and voltages have been increased, then even more so.

I'm suspicious that the VRM section that provides Vcore could be dying, maybe starting to fail short, which could also cause a sag on the +12V line which powers it.
The popping sound could be repeated cycling of a power line that's overloaded.

I'm just speculating though. Try to measure as many voltages as you can and look for anything that's unstable or incorrect.

Nexxen wrote on 2025-05-11, 09:21:

Pics of the heatsink, pls.
Warm it and remove the lid using a fine metal thread. Requires a vice. You can try without warming first.

Below is the heatsink pic. I tried using a different P4 processor and another heatsink. However, no dice.

asdf53 wrote on 2025-05-11, 09:24:

While the board is running, measure the voltage between the MOSFET output pin and ground. It should be close to the expected core voltage of the CPU.

Ok. I measured the MOSFET voltages and it was at 0.8V for both the ICs.

shamino wrote on 2025-05-11, 09:44:

I'm suspicious that the VRM section that provides Vcore could be dying, maybe starting to fail short, which could also cause a sag on the +12V line which powers it.
The popping sound could be repeated cycling of a power line that's overloaded.

Any idea of the VRM in this board? I did notice an ic named HIP5302CB above the 4 pin power connector. However, I was unable to find any datasheet on it.

stealthjoe wrote on 2025-05-11, 09:49:

asdf53 wrote on 2025-05-11, 09:24:

While the board is running, measure the voltage between the MOSFET output pin and ground. It should be close to the expected core voltage of the CPU.

Ok. I measured the MOSFET voltages and it was at 0.8V for both the ICs.

I'm not familiar with the Pentium 4, but that seems quite low. Are you measuring the right FET and pin? I meant the FETs next to the CPU socket, where the two large inductor coils are. The vcore pin is shared between the source of one FET and the drain of an adjacent FET. Try the right pin of Q281 or the middle pin of Q282.

You can differentiate the pins as follows: Motherboards use mostly n-channel FETs, the large tab connected to the middle pin is often the drain, the pin that is connected to another large plane is the source, and the pin that looks a bit isolated (not connected to any large plane) is the gate, sometimes next to a little resistor, connected to the controller IC.

Measuring the FETs with a multimeter in diode mode can be tricky when they're still connected to the board, I often get readings that are all over the place. I'm not very knowledgeable about this, but I believe this could be due to leakage current in old capacitors, so some FET might appear shorted when they are actually fine.

asdf53 wrote on 2025-05-11, 11:07:

I'm not familiar with the Pentium 4, but that seems quite low. Are you measuring the right FET and pin? I meant the FETs next to the CPU socket, where the two large inductor coils are. The vcore pin is shared between the source of one FET and the drain of an adjacent FET. Try the right pin of Q281 or the middle pin of Q282.

You can differentiate the pins as follows: Motherboards use mostly n-channel FETs, the large tab connected to the middle pin is often the drain, the pin that is connected to another large plane is the source, and the pin that looks a bit isolated (not connected to any large plane) is the gate, sometimes next to a little resistor, connected to the controller IC.

Measuring the FETs with a multimeter in diode mode can be tricky when they're still connected to the board, I often get readings that are all over the place. I'm not very knowledgeable about this, but I believe this could be due to leakage current in old capacitors, so some FET might appear shorted when they are actually fine.

For better clarity, I have marked the ones I tested in the image below. The ones in red measured 0.8V. Within the blue circle one measured 1.5V and the other 10V. I then proceeded to measure the other ICs in the same order above ( left of cpu socket) and they all measured similar values (1.5V and 10V).

Btw, the measurement was done for all the MOSFETs assumption as N MOSFET i.e. right pin is Source, center pin and the ground plane is Drain. The left pin is Gate.

If the source pin of the upper one and the drain pin at the lower one in the blue circle are at 1.5V, and that's the Vcore of your CPU, it should be good. But you're doing this with the CPU installed and running, right?

The lower ones that produce 0.8V, no idea. I'd try and see where their source pin leads to, and figure out if that's the voltage the part needs. Could these lead to the Northbridge? If so, that seems too low.

If that seems to be suspicious, do as Imperious suggested above, heat up the caps around the northbridge with a hairdryer and see if that makes the board POST more reliably.

You could also test the voltage of the drain pin and the gate pin of the red circled FETs. Drain pin should be a supply voltage such as 5V or 12V, and the gate pin should be slightly higher than the supply voltage.

I am also thinking if it is something to do with the RAM slots. I used to face intermittent problems like this earlier. Then, I would remove and reseat the RAM. The board then posted. The board used to post only in that slot. I am afraid that both the RAM slots have failed. Is there any way to find out if the RAM slots are actually working or not?

Hey , when you have ram issues your diagnostic card gives you codes like C0 or C1 . I noticed atx 20 pin connector has burn signs , does its because of soldering attempt ? If not it might be clue. Other than this vrm side seems OK. This issue might be failed chipset also .

ciornyi wrote on 2025-05-12, 07:02:

Hey , when you have ram issues your diagnostic card gives you codes like C0 or C1 . I noticed atx 20 pin connector has burn signs , does its because of soldering attempt ? If not it might be clue. Other than this vrm side seems OK. This issue might be failed chipset also .

Hi. Unfortunately, the only code I get is 0000 with the diagnostic card. It doesn't even change unlike in other cases where the card cycles through many numbers while booting. Reg the burnt part, yes. It was a soldering job to replace two caps above the ATX connector. However the board didn't post and had the same issue even before changing them.

The northbridge and southbridge both feel cold and not warm enough even several minutes after booting up the board. So I am suspecting if these have failed or maybe not getting the voltage.

This board and for that matter any working Socket 478 is extinct in my locality.

When the board is powered up, do the CPU and the northbridge stay cold? The southbridge as well? I believe the sequence goes something like this:

- Power button is pressed, PSU signals PWR_OK
- Super I/O chip senses this and propagates power OK
- Southbridge senses this and turns on the northbridge (PCIRST# goes high)
- Northbridge senses the reset signal, turns itself and the CPU on

At one point, this chain seems to be broken. One thing I'd try is see if the northbridge receives a stable 1.5V at its power pins/balls.
And try if you can somehow test the state of the PCIRST signal with your multimeter. The pins of the chipset are not directly accessible because it's BGA, so it could be tricky to find it.

stealthjoe wrote on 2025-05-11, 13:56:

For better clarity, I have marked the ones I tested in the image below. The ones in red measured 0.8V. Within the blue circle one measured 1.5V and the other 10V. I then proceeded to measure the other ICs in the same order above ( left of cpu socket) and they all measured similar values (1.5V and 10V).

The attachment IMG_20250511_192158.jpg is no longer available

Btw, the measurement was done for all the MOSFETs assumption as N MOSFET i.e. right pin is Source, center pin and the ground plane is Drain. The left pin is Gate.

10V is odd to find on a motherboard.
Is that coming from the 12V rail? With the board running, check your 12V (and all the other supply voltages while you're at it).

I re-measured the MOSFET voltage outputs and below are the findings:

For the above MOSFETs the output voltage measured as 2.3V at the Drain for the bottom one and 3.2V at the Drain for upper one.

Here, the MOSFETS along the left side (towards the output ports) all measured 11.96V and on the right side measured 1.5V.

Intermittent issues are (almost) never a MOSFET problem.

Given the state of the original CPU (with the lid / heat spreader removed), I can't help but suspect lots of force was used at some point to remove the heatsink from the CPU. Did someone use thermal epoxy on that? If so, I wouldn't be surprised. See if you can use a sharp flat heat screw driver to separate the heat spreader from the heatsink. If you can't, then that explains why it's stuck on there... and again, probably at some point, a lot of force was used to remove that CPU from the motherboard without the latch being open. With this comes the risk of damaging both the CPU socket itself and the BGA under the CPU socket - something which I have done myself once as well (on a LGA775 mobo). Flex it enough times while installing the cooler/heatsink, and that's most likely to be the issue here. Of course, it could also be bad contacts in the RAM slots or the CPU socket itself too... though at this point, these should have cleared when you tested different CPUs and RAM. So I suspect these are not the issue. It could also be dirt/dust buildup causing issues... though again, I think this is less likely to be it. In any case, it probably won't hurt to have the motherboard washed with water and dish soap. Then let it dry for a few days under the sun or in a breezy place. But if that doesn't help, then I highly suspect the CPU BGA is damaged somewhere. The stock Intel heatsink retention mechanism for P4 is already bad enough (at least if it uses the cooler design with double-latches with lobes on each side) as many of these tend to warp the board under the CPU socket. Most s478 boards that became intermittent and died "for not apparent reason" were a case of this, IME, bad caps excluded of course.

If nothing seems to work, remove the CPU heatsink plastic retainer from the board, try installing another CPU, and put any heatsink on it that fits (can be LGA775 heatsink, socket 754/939/AM2/AM3, etc.) without clamping it down. Now try powering on the board. Can you get it to boot? If not, how about if you manually apply some pressure with your hand on the heatsink and then try to power on the PC while still applying pressure? Can you get the board to boot then? If yes, you most certainly have a bad BGA in the CPU socket. If not, then... well, there's still no guarantee that's not the issue. Sometimes a bad BGA joint may not get restored simply by applying pressure. But with 1 out of 3 boards, I find that it does, so why I recommend it as something to try.

As for bad MOSFETs... again, that just can't be it, really. Bad MOSFET usually means shorted MOSFET. Most MOSFETs tend to either switch (chop) or linearly regulate a high(er) voltage rail down to a lower voltage. In the case of the CPU MOSFETs, you get 12V as input and 1.5V as output to the CPU core. If an upper MOSFET was bad in the CPU VRM, you'd get 12V going to the CPU core, and that usually a) causes an over-load / short-circuit protection on the PSU to kick in and b) kills the CPU (though P4's tend to be pretty tough and hard to kill IME, so point b is less likely here.) And if it was a lower MOSFET shorted, you'll have a straight short-circuit to ground on CPU core... so again, a bad MOSFET in the CPU VRM is not possible here given the symptoms you describe. A short MOSFET elsewhere is also unlikely. For example, the RAM VDIMM voltage (2.5V for DDR, 3.3V for SDR): for DDR, the 2.5 or 2.6V supply is typically generated by a linear regulator from the 3.3V rail on the PSU. If the MOSFET that does this goes bad, it will send full 3.3V to the DDR RAM module... making the chips overheat and/or smoke. That didn't happen in your case, so you can be sure this was not the issue. Same goes for SDR RAM: it takes 3.3V for the ICs, which is generated by a linear regulator from the 5V rail typically.

momaka wrote on 2025-05-14, 15:38:

Intermittent issues are (almost) never a MOSFET problem.

Given the state of the original CPU (with the lid / heat spreader removed), I can't help but suspect lots of force was used at some point to remove the heatsink from the CPU. Did someone use thermal epoxy on that? If so, I wouldn't be surprised. See if you can use a sharp flat heat screw driver to separate the heat spreader from the heatsink. If you can't, then that explains why it's stuck on there... and again, probably at some point, a lot of force was used to remove that CPU from the motherboard without the latch being open. With this comes the risk of damaging both the CPU socket itself and the BGA under the CPU socket - something which I have done myself once as well (on a LGA775 mobo). Flex it enough times while installing the cooler/heatsink, and that's most likely to be the issue here. Of course, it could also be bad contacts in the RAM slots or the CPU socket itself too... though at this point, these should have cleared when you tested different CPUs and RAM. So I suspect these are not the issue. It could also be dirt/dust buildup causing issues... though again, I think this is less likely to be it. In any case, it probably won't hurt to have the motherboard washed with water and dish soap. Then let it dry for a few days under the sun or in a breezy place. But if that doesn't help, then I highly suspect the CPU BGA is damaged somewhere. The stock Intel heatsink retention mechanism for P4 is already bad enough (at least if it uses the cooler design with double-latches with lobes on each side) as many of these tend to warp the board under the CPU socket. Most s478 boards that became intermittent and died "for not apparent reason" were a case of this, IME, bad caps excluded of course.

If nothing seems to work, remove the CPU heatsink plastic retainer from the board, try installing another CPU, and put any heatsink on it that fits (can be LGA775 heatsink, socket 754/939/AM2/AM3, etc.) without clamping it down. Now try powering on the board. Can you get it to boot? If not, how about if you manually apply some pressure with your hand on the heatsink and then try to power on the PC while still applying pressure? Can you get the board to boot then? If yes, you most certainly have a bad BGA in the CPU socket. If not, then... well, there's still no guarantee that's not the issue. Sometimes a bad BGA joint may not get restored simply by applying pressure. But with 1 out of 3 boards, I find that it does, so why I recommend it as something to try.

The board was obtained locally in 2020 and the initial state inside the case was pathetic. There was excessive cobweb as well as spiders moving around. I cleaned up the excessive dust but didn't remove the board due to fear of making things worse. The board was acting strange at times with similar symptoms such as not posting and getting the faint beep sounds. This used to happen occassionally since I got it. I was also getting hard disk not found error. So, once I reconnected the hard disk and reseated the RAM it used to post fine.

If I can recollect, the board stopped posting once I removed the other AGP card. This card felt a little firm to come out and I pulled the small latch at the end of the AGP slot to try to loosen it. However the card came out with quite a bit of resistance and not smoothly. This caused the board to go up and drop down albeit not with much force. I am thinking if this caused some solder joints to break.

I did try to start the board by loosening the heatsink clamps, but the same result. I am trying to understand the difference between applying manual pressure on the cpu vs the pressure applied by the heatsink clamps.

Ah, looks like the damage might have been done before you even got the PC then.
Removing the video card mostly likely just "upset" whatever was already broken (could be, and I think this is the most likely case, bad BGA either under CPU socket or possibly Northbridge.) So don't feel bad about it.
I have a few boards like this. Flexing them around can sometimes get them to POST or even work fine for a few sessions. They are just not reliable in the long-term... and that's that. I could re-work them, but it would really be too much effort for boards that aren't really worth it.

stealthjoe wrote on 2025-05-15, 04:32:

I did try to start the board by loosening the heatsink clamps, but the same result. I am trying to understand the difference between applying manual pressure on the cpu vs the pressure applied by the heatsink clamps.

The heatsink clamps always apply the same pressure... and thus warp the board (under the CPU socket) by the same amount. Manual pressure allows you to vary this, so you can see if anything changes.
On that note, also recommended to manually try applying some pressure on the Northbridge heatsink too (prior to pressing the power button, and keeping the pressure), just to see if somehow the BGA of that wasn't damaged either.

And give the motherboard a wash. It won't hurt things, but definitely could improve them. I've had intermittent dusty/dirty components like this come back to life after a full wash and proper drying.

Update:

I gave a good bath to the motherboard using soap water after removing the battery, CPU and RAM. Then dried it for over 24 hours. After that I further dried it using a hot air gun at low setting on all the slots as these are the places where drying takes eternity. Also applied hot air on high setting over the northbridge and southbridge. Connected the board and tried to start, but still no luck. Later I noted that there was a broken trace near one of the caps I replaced (close to the ATX connector). This damage was visible only after washing the board as there was some residual flux over it. I somehow managed to fix it after a few setbacks yesterday.

After soldering the trace properly, I connected the board and tried starting. This time, the board sprung to life. There was output on the monitor and it was such a delight to see the Intel splash screen after missing it for many moons (actually for only 10 days 😜). Hooorrraayyyy!!!! 😁 😁 😁 . Tried running some games and it worked good. However, I had to change the PSU as it was causing the system to shutdown immediately after entering Win 98. Now the build seems fine and stable. Thank you all for your help and inputs. It is my birthday today and I am happy that things worked favorably this time.