I don't have that board, but you can't miss if you replace these caps ...

... with 6.3V rated lytics of the largest can size and lowest ESR that will fit. Like these series:

Chemicon KZE / KZG / KZH / KZM / KZN

Panasonic FM / FR / FS

Rubycon ZL / ZLG / ZLH

Hi, thanks for the tips and help. I only found a few blurry photos online; these are the most detailed ones—you can see several 10V and 16V capacitors in them, but I’m not sure if they’re the original ones or if they’ve already been replaced.

That's a Socket 370 board, from the times prior to the introduction of the "P4" connector and +12V feed for the CPU.

The highest voltage any cap around the CPU (and most probably across the whole 370 board) will experience is +5V. Thus 6.3V rated caps will be fine (10V if we're paranoid).

Sometimes OEM use caps of weird ratings for reasons of their own. Then confused enthusiasts recap with even weirder ratings, see here.

Superb advice from PCBONEZ on recapping Socket 370 (TUV3X) mobo here.

6.3V rated Rubycon ZL around the CPU in a Socket 370 TUV4X pic from The Retro Web:

The datasheet for the used PWM controller (ISL 6524CB) specifies 3x1000uF electrolytic capacitors for Vcore (1.825V max) and 1x for the GTL bus (1.2V). The AGP bus and MCH use 560uF capacitors—Soltek originally used 1000uF 6.3V GSC capacitors. I think that GSC capacitors were originally used on the Vcore branch as well, and they were notorious for swelling and leaking electrolyte. According to some user comments on forums, Soltek even sent out a set of new ones free of charge for replacement.
I think the 6.3V capacitors will be sufficient, but the mounting area is designed for 12mm types, so I could also use 10V ones with higher capacitance and RC values. Among the ones I have available here are Nippon Chemicon 1500uF 16V KZH capacitors, and instead of the smaller ones, I’d use 1000uF 10V KZM capacitors—would that work?

Ydee wrote on Yesterday, 08:48:

If anyone has this board, could you share the specifications of the main capacitors? The photos online aren't very clear, and it seems that Soltek used different variants (judging by the colors) depending on what they had in stock at the time.

I wouldn't worry too much about the ESR spec of the capacitors. Ultra-low ESR electrolytic capacitors were not yet around when Pentium 3 boards were mainstream. So going with any of the following brands and series should be just fine:

Rubycon: ZL, ZLH, ZLQ, ZLG
United Chemicon: KZE, KZH, KZM, KY, KYB... and LXY/LXV/LXZ probably OK too for anything but the CPU V_core output (though I think these will work too, in a pinch)
Panasonic (Matsushita): FR, FS, FM, FK... or new-old-stock (NOS) FJ/FJS
Nichicon: HW, HV, HE

As TELVM mentioned, this board uses pretty much 3.3V and 5V power for everything, so you can go with 6.3V and 10V caps on just about any spots. Just beware of any 16V-rated caps around the 20-pin ATX connector and possibly around the AGP slot, as those might actually have a 12V rail running through them. But outside of those edge cases, it's fine to use 6.3V and 10V caps. I usually do a quick "cap-map" with a multimeter by checking which caps' positive leads are connected to which power rail on the ATX connector - i.e. the 3.3V, 5V, and 12V rails. For 3.3V rails, I usually use 6.3V caps and for the 5V rail 10V caps... but won't hesitate to mix-n-match if my cap stock goes low on one or the other.

Hi momaka, thank you for tips. As for the AGP, I'm sure the capacitors near it are the original ones—1000uF 6.3V—so a 10V replacement will definitely suffice. Next to the ATX connector, there's one larger one, also 1000uF 6.3V, and two small ones, 22uF 25V—but I won't be replacing those.
Cap-map is, of course, a great idea!

By the way, when you find yourself thinking that 6.3 V is a bit on the low side and decide to fit a bigger capacitor – say, a 10 V one – don’t forget that an electrolytic capacitor is, by its very design, simply a roll of foil.
In other words, it was inductor.
And switching regulators really don’t need any extra inductance.
What’s more, it significantly worsens noise filtering anywhere in the circuit.
The same goes for capacitance. You can end up with a bump in the impedance curve. This happens when there aren’t enough ceramic capacitors,and a single large electrolytic capacitor simply cannot cope with the task purely because of its physical size.

@Ydee

These are the values. Compiled from both TELVM's photo and post #3.
C36, C37, C39, C43 - 2200uF 6.3v
All the green asterisk caps by TELVM - 1000uF 10v (according to the original blue GSC from post#3.)

shevalier wrote on Today, 12:15:

By the way, when you find yourself thinking that 6.3 V is a bit on the low side and decide to fit a bigger capacitor – say, a 10 […]
Show full quote

By the way, when you find yourself thinking that 6.3 V is a bit on the low side and decide to fit a bigger capacitor – say, a 10 V one – don’t forget that an electrolytic capacitor is, by its very design, simply a roll of foil.
In other words, it was inductor.
And switching regulators really don’t need any extra inductance.
What’s more, it significantly worsens noise filtering anywhere in the circuit.
The same goes for capacitance. You can end up with a bump in the impedance curve. This happens when there aren’t enough ceramic capacitors,and a single large electrolytic capacitor simply cannot cope with the task purely because of its physical size.

So, what now? There are no photos of the board on Retroweb, let alone one in its original condition with the original capacitors, and I’m afraid that, given the “capacitor plague” at that time, very few people still have it in its original condition. From the photos online, I can’t tell what’s original and what’s been replaced with different specifications. That’s why I’m asking here what would be best.
Right now, there are those replaced 1500uF 6.3V Jamicon capacitors everywhere, and the original 1000uF 6.3V GSC capacitors—I can replace those with exact replacements, because those are definitely original. The board works, but it's not entirely stable, so I want to recap it.

PcBytes wrote on Today, 12:38:

@Ydee […]
Show full quote

@Ydee

These are the values. Compiled from both TELVM's photo and post #3.
C36, C37, C39, C43 - 2200uF 6.3v
All the green asterisk caps by TELVM - 1000uF 10v (according to the original blue GSC from post#3.)

Thank you, PcBytes - post#3 It's from the internet, and I can't tell if it's the original configuration or a recap. The 12mm diameter for the VRM suggests that it could be either a higher voltage or a higher capacitance, but I can't say for sure.

Another photo:

1500uF 6.3v, from the same source. Note the 6v3 text. Same goes for C43 apparently.

C36,37 and 39 are 2200, as I mentioned before however, you can see the 22 on the Rubycons in the photo.

Yeah, I see this and for those black EPCOS we can suppose they are for NB voltage, so 6,3V too good enough. Thank you all, friends!