On these old mainboards, without step-down regulators, you can use "general purpose" electrolytics, or Low-ESR types.
It doesn´t matter, because they are just local buffers compensating transients and doing some filtering etc.

What was the test frequency when you measured 0.5 Ohms ESR? I doubt they have low ESR like that.

The Gigabyte 586HX-2 does have a step-down regulator, so all the electrolytics behind the inductor and those around the CPU socket should be "regular" low ESR like Rubycon ZL, ZLH, ZLS.

majestyk wrote on 2023-03-11, 20:26:

On these old mainboards, without step-down regulators, you can use "general purpose" electrolytics, or low-esr types.
It doesn´t matter, because they are just local buffers compensating transients etc.

What was the test frequency when you measured 0.5 ohm ESR? I doubt they have low ESR like that.

Actually, the GA-586HX2 even has a buck PWM controller for the CPU voltage (HIP6008CB) but I couldn't find anything speaking against low-ESR caps in the datasheet.

But apart from that, I just realized that stupid me was measuring the caps with leads which added to the ESR measurement. With only short legs soldered to the caps, I measure only about 0.1 to 0.2 Ohms of ESR, so that now makes more sense.

I've bought caps with 0.045 Ohms (1000uF, 10ZLH1000MEFC8X16) and 0.075 Ohms (330uF, 25ZLJ330M8X11.5), so easily half the original value but I guess not many circuits can get thrown off by that anyway...

majestyk wrote on 2023-03-11, 20:26:

The Gigabyte 586HX-2 does have a step-down regulator, so all the electrolytics behind the inductor and those around the CPU socket should be "regular" low ESR like Rubycon ZL, ZLH, ZLS.

Nope, only three types of electrolytic caps. Two different brands 1000uF 10V and one type 330uF 25V, that's it, they're all the same and they measure all around 0.1-0.2 Ohms ESR with my cheap device which I have no idea which frequency it uses to measure ESR 🤣

- Chemicon
- Kemet
- Nichicon
- Panasonic
- Rubycon
- Vishay
Any of these brands have standard and low-ESR capacitors and they are good . You can pick according to price / availability in your area . I tend to stick with the original layout of the board (replace standard with standard and low ESR with low ESR ) but if i have standards in the buck converters (VRM's) output i replace them with low ESR .

If you really want standard 1000uF/10V caps with 8mm diameter you can get these Rubycons from mouser (PX series):

https://www.mouser.com/ProductDetail/Rubycon/ … RaDtDdz0Q%3D%3D

You can also get Chemi-con LXZ or Panasonic FC, which are low ESR but not so low as to cause any issues.

If possible, try to source datasheets for the original caps, that will allow you to determine whether they are low ESR or not. Those chinese ESR meters are not very reliable I think.

those Chinese meters are on point, measuring esr is not rocket science.
I wouldnt bother with any expensive special caps for socket 7, CPUs maxed out under 20W back then. Random $0.1 Elite caps will outlive you or your passion for retro 😀

ESR depends on uF and voltage and ESR is measured with 120Hz frequency if otherwise stated in the datasheet of given capacitors. In Vcore and Vio planes, they have to be low esr regardless, *not* general purpose. A good esr of 330uF is about .2 or so for 6.3V to 10V. As uF goes up eg 470uF and up eg your example 1000uF, the good ESR goes down is a rule to look for, 6.3V and 4V are good subs and should be less than ESR of 0.1 to 0.05.

I was working with motherboard capacitors for repair and know what they should be. As you parallel 2 or more, in that power plane, ESR goes lower, current ripple rating goes up, ESR/n where number of capacitors of like.

Low esr is a requirement on pentium and later due to need for high current ripple rating which low ESR capacitors are built for.

Cheers,

nemail wrote on 2023-03-11, 20:15:

but I also know that under certain circumstances like with some LDO voltage regulators and some types of switching voltage regulators and even more components, low-ESR caps can lead to instabilities.

It came from audiophiles. LDOs at low currents with LOWESR capacitors lead to an increase in noise, sometimes by 10 dB or more.
In power circuits, a low-resistance load serves as a damper that disrupts ringing.
The voltage ripple is equal to the sum of the noise and the product of the ESR * ΔI.
With current changes of ten amperes, noise no longer plays a role.

For yellow PCB, standard red-marked polymer capacitors are better suited - Fujitsu, Wurst, Lelon

The range of viable cap esr value depends on the design of the regulator feedback loop and the response characteristics of the regulator itself.
https://www.ti.com/lit/an/slva115a/slva115a.pdf

High current pwm regulators will require ultra low esr caps (realistically polymers these days as ultra low esr Al-EL caps like Nichicon HN, NCC KZJ are difficult to find) to achieve decent ripple values. You see this a lot in Athlon 64 and P4 cpu vrms. You can get a lot of mileage out of gp caps like Nichicon PW for more sedate designs.

gdjacobs wrote on 2023-03-14, 01:34:

The range of viable cap esr value depends on the design of the regulator feedback loop and the response characteristics of the regulator itself.
https://www.ti.com/lit/an/slva115a/slva115a.pdf

Re: The graphics card repair attempt megathread.

Stable operational area its not а dot, its big ... area.
The main rule is that the smaller the ESR - the larger the capacitance.

shevalier wrote on 2023-03-14, 05:06:

Re: The graphics card repair attempt megathread. https://www.vogons.org/download/file.php?id=159502&mode=view Stable operational […]
Show full quote

gdjacobs wrote on 2023-03-14, 01:34:

The range of viable cap esr value depends on the design of the regulator feedback loop and the response characteristics of the regulator itself.
https://www.ti.com/lit/an/slva115a/slva115a.pdf

Re: The graphics card repair attempt megathread.

Stable operational area its not а dot, its big ... area.
The main rule is that the smaller the ESR - the larger the capacitance.

Agreed, it's not a highly sensitive parameter unless you're making a huge change in cap. A close match is never wrong, but you can usually get away with a much looser approximation.