Which board that is and does it have official support for coppermine and what kind of voltages it can provide? This is what matters and not just some bios compatibility.

That CPU is relatively high power to begin with, little over 30W or so at 1,45V-1,5V. Coppermines have 1,6-1,75V and Katmais 1,93-2,1V. If your mobo voltage regulator can’t go lower than 1,8V for example, like many 440BX boards can’t, you are blasting that CPU with too high voltage and it draws significantly more power, something like 40W. That will put the that VRM under strain and it is no wonder that regulators start to heat up, it wasn’t designed for those kind of loads and it is easily 20% more than the highest Katmai has.

Tayeh caps are bottom tier plague era junk, right down there with GSC, and I would replace them immediately no matter how they look. I don’t think they are the source of your heating problems. Without you providing more information, my bet it is that you have relatively high power CPU which you are overvolting like crazy and powerdraw is out of spec for that board.

You can also install one of those 9x9mm heatsinks using a thin double-sided tape on the regulators.

dominusprog wrote on Yesterday, 17:11:

You can also install one of those 14x14mm heatsinks using a thin double-sided tape on the regulators.

If the board OP is using can’t deliver low enough voltages for Tualatin, no heatsink will take away the load that those regulators have. Or remove the overvolting the CPU. This can pose a serious risk to hardware especially when used longer periods of time. I’d recommend stopping using that CPU with the board if it can’t be adjusted low enough or OP has something like powerleap adapter that can set CPU voltages to correct levels and which VRM can most likely also handle.

CharlieFoxtrot wrote on Yesterday, 17:20:

dominusprog wrote on Yesterday, 17:11:

You can also install one of those 14x14mm heatsinks using a thin double-sided tape on the regulators.

If the board OP is using can’t deliver low enough voltages for Tualatin, no heatsink will take away the load that those regulators have. Or remove the overvolting the CPU. This can pose a serious risk to hardware especially when used longer periods of time. I’d recommend stopping using that CPU with the board if it can’t be adjusted low enough or OP has something like powerleap adapter that can set CPU voltages to correct levels and which VRM can most likely also handle.

Well, I had the same issue of overheating of the VRM on my Slot 1 motherboard and the 800MHz 133FSB CPU which is supported by the board.

dominusprog wrote on Yesterday, 17:25:

CharlieFoxtrot wrote on Yesterday, 17:20:

dominusprog wrote on Yesterday, 17:11:

You can also install one of those 14x14mm heatsinks using a thin double-sided tape on the regulators.

If the board OP is using can’t deliver low enough voltages for Tualatin, no heatsink will take away the load that those regulators have. Or remove the overvolting the CPU. This can pose a serious risk to hardware especially when used longer periods of time. I’d recommend stopping using that CPU with the board if it can’t be adjusted low enough or OP has something like powerleap adapter that can set CPU voltages to correct levels and which VRM can most likely also handle.

Well, I had the same issue of overheating of the VRM on my Slot 1 motherboard and the 800MHz 133FSB CPU which is supported by the board.

If it is officially supported, then it is most likely operating well within the design parameters and thus not overheating. Unless it is a bad design and regulators indeed go to 90-100C or so.

A 440BX boards aren’t designed for a Tualatin OP has and if it cant go low enough voltage, it is almost certain that at 40W or more the regulator is seriously overloaded. As is the CPU.

I will further clarify.
The board voltage regulator have been swapped with one capable of delivering correct voltage down to 1.4v.
The slotket are high quality, either Asus or QDI, both capable of voltage settings.
The board works just fine with the Celeron and can happily overclock it.
CT-6BTM with modded bios.

Is it worth changing the caps with some higher voltages ? Would this mitigate overheating?

What is getting hot exactly, caps? because transistors are good up to 170C 😀 and can do at least 10A each at 100C

On a Pentium 3 motherboard with fully functional capacitors (actually working ones, not just "looking good" like people love to say), the capacitors themselves will stay cool.
The MOSFETs will be noticeably warm, and the inductor can get pretty hot (which is totally normal).

shevalier wrote on Today, 04:05:

On a Pentium 3 motherboard with fully functional capacitors (actually working ones, not just "looking good" like people love to say), the capacitors themselves will stay cool.
The MOSFETs will be noticeably warm, and the inductor can get pretty hot (which is totally normal).

Not to this particular case specifically, but this got me thinking.

Obviously caps aren't heating significantly (unless they are shorted 🤣), but it would be interesting to calculate how much higher ripple would heat up regulators. I mean, if input caps would have gone bad and ESR shot up, wouldn't higher ripple as a consequence manifest to higher temps on regulators? Or is it this obvious?

rasz_pl wrote on Today, 03:35:

What is getting hot exactly, caps? because transistors are good up to 170C 😀 and can do at least 10A each at 100C

This. Hot as "it feels burning to my fingers" is meaningless, as you get that sensation at about 50C. What matters is the manufacturer spec and actual temperature measurement.

No one has yet asked to check if the CPU is really powered with the intended voltage. This is the first thing to check.

The voltage can be wrong because SW reasons, but also for HW reasons.

I discovered, the hard way, that a not properly seated CPU can lead to a mismatch between the contacts that inform the MB what kind of voltage the CPU requires.

A coworker assembled a PC (a BX mobo + a legendary Celeron 300 Mendocino) then handed it to me for installing the OS and other SW. The machine was still open and after 1/2 hour i started to smell a strange odor, not the magic smoke odor, but still strange. Stupidly I touched the heatspreader, and i got burned, I mean not not just the feeling, the thing was REALLY hot, surely well above 100 degrees, I have still the mark on my finger almost 3 decades later.

Believe it or not the PC was still running perfectly, and AFAIK it's still alive and kicking in 2026, running for decades @504MHz (with the CPU properly seated, obviously), at 504MHz the thing was actually faster than the richer brother Pentium II that cost 5x the price of a Mendocino.

Back to the contacts... Some dirty contacts may lead to a similar scenario. (IIRC) the voltage selection contacts are located on the extreme side of the connector, so a little misalignment is amplified by the position.

CharlieFoxtrot wrote on Today, 04:16:

shevalier wrote on Today, 04:05:

On a Pentium 3 motherboard with fully functional capacitors (actually working ones, not just "looking good" like people love to say), the capacitors themselves will stay cool.
The MOSFETs will be noticeably warm, and the inductor can get pretty hot (which is totally normal).

Not to this particular case specifically, but this got me thinking.

Obviously caps aren't heating significantly (unless they are shorted 🤣), but it would be interesting to calculate how much higher ripple would heat up regulators. I mean, if input caps would have gone bad and ESR shot up, wouldn't higher ripple as a consequence manifest to higher temps on regulators? Or is it this obvious?

Capacitors with a high ESR heat up quite a lot – up to 50+ degrees.
This is particularly true of those located in front of the VRM, as there are usually only a few of them.
And when they can no longer fulfil their function, the controller extends the pulses, which leads to the whole system heating up more.

CharlieFoxtrot wrote on Today, 04:18:

rasz_pl wrote on Today, 03:35:

What is getting hot exactly, caps? because transistors are good up to 170C 😀 and can do at least 10A each at 100C

This. Hot as "it feels burning to my fingers" is meaningless, as you get that sensation at about 50C. What matters is the manufacturer spec and actual temperature measurement.

Firstly, this applies to modern semiconductors, modern casings and modern solders.
Back in those days, 105 degrees was already considered a cracking achievement.

Secondly, this is the die temperature; the temperature of the package will be 20–30 degrees lower.
Therefore, if you cannot keep your hand on an old MOSFET for even a short while (65–70 degrees, as far as I recall), then the chip is already dangerously close to failure.

This isn’t the RTX5090 from the year before last; electronics from those days weren’t designed to handle such temperatures.

shevalier wrote on Today, 06:12:

Firstly, this applies to modern semiconductors, modern casings and modern solders.
Back in those days, 105 degrees was already considered a cracking achievement.

you are thinking logic chips, not power electronics
this is for the particular transistors as mounted on this board https://www.kontest.ru/datasheet/philips/phb55n03lt.pdf
modern ones go up to 11.. erm 200C of safely operating temp

shevalier wrote on Today, 06:12:

Firstly, this applies to modern semiconductors, modern casings and modern solders.

You are too confident in modernity, actually "modern solder" was the main source of electronics failures for a decade, in the 2000s, while electronics from 60/70/80/90 never suffered big problems, unless some unskilled worker did a poor soldering job, during the production phase.

Secondly, this is the die temperature; the temperature of the package will be 20–30 degrees lower.

Don't apply the logic applicable with CPUs to power transistors, in the latter the chip is bonded on the support metal, which means very low thermal resistance, which means there is no way to have 30 degrees of difference between the chip and the transistor externals.

electronics from those days weren’t designed to handle such temperatures.

Most of the improvements are actually related to the voltage drop out, which means less heat generated for a given current and / or frequency (or more current managed for a given power dissipation), higher operating frequencies/voltages and alike. Improvements on the operating temepratures are there but aren't huge. Unless you look to completely different, more modern, semiconductors like Silicon Carbide or Gallium Nitride, that can manage way higher temperatures than Silicon.

rasz_pl wrote on Today, 06:37:

this is for the particular transistors as mounted on this board https://www.kontest.ru/datasheet/philips/phb55n03lt.pdf

How can these semiconductors be used at such temperatures with such derating?

shevalier wrote on Today, 07:32:

rasz_pl wrote on Today, 06:37:

this is for the particular transistors as mounted on this board https://www.kontest.ru/datasheet/philips/phb55n03lt.pdf

How can these semiconductors be used at such temperatures with such derating?

I don't know, but to me for example 100C looks still completely usable unless I'm reading something wrong. You obviously need to take derating into account with the design, 100% doesn't happen in real life applications like pretty much ever.

shevalier wrote on Today, 07:32:

How can these semiconductors be used at such temperatures with such derating?

Exactly as I said in my first post - at 10A actually its 38A at 100C, giving whole regulator at least 30W _even more_ of safe capacity.

shevalier wrote on Today, 07:32:

How can these semiconductors be used at such temperatures with such derating?

You need to understand how companies think, especially chinese ones.

Often an over specced semiconductor is way cheaper than the additional cost of an headspreader (and the manual labor to apply it).

This usually doesn't work for an hobbyist residing in the west, who buy semiconductors in units, or dozens, possibly from a local physical shop, but works very well for a chinese company close to the other chinese company that manufacture semiconductors, and buy components in thousands.

In short any, serious, discussion involving math and specs, needs to start from the actual components used, the actual current involved, temps and so on.

Listen, we’re talking about the year 2000.
There were still 20 years to go before the release of the Cadence Celsius Thermal Solver.
There were no elaborate thermal or power calculations involved.
They simply built in a margin of about 100 per cent, and that was it.
ATI also fell for the semiconductor manufacturers’ claims about 100500 degrees Celsius in the Radeon 9700/9800.
As a result, they had to remove the fan controller from the board and glue heat sinks onto the MOSFETs in the reference design.

That’s why there can be no excessive heating of power components on a Pentium 3. With the exception of AGP linear regulators, but the motherboard manufacturers only have themselves to blame for coming up with such nonsense.