I am a big believer in starting with the obvious first. This power supply does seem very small compared with the power hungry overclocked Athlon and two video card you have stuffed in there. I don't even run a DOS box with less than a solid 300+ supply. A set up like this screams 350-450 if you can swing it. Hope you threw in a case fan or two to push out all of that toasty goodness!

Yep, you're probably right! I have a spare Seasonic 450W that I will try. Case fans will get installed as well as a new Northbridge cooler (the Abit board has active cooling with a tiny, tiny heatsink). For now, I'm running the case open. I know that the basic setup minus the Voodoo 2 is stable, though.

BTW, the Athlon is not overclocked, it's slightly underclocked 😉 I've used the pencil trick to up the multiplier to 13, otherwise it'd run at 1GHz. 133MHz FSB processor in a 100MHz board.

Yeah that Athlon is one of the most power consuming models. It can pull up to 70W. Definitely don't want all of that hardware on a 235W PSU.

Also take note that a KT133 board will pull almost all power from 5v. What is the 5v rating on that 450W? I suspect it will be fine though.

Didn't know these boards draw processor power from the 5V line, these days everything's about the 12V line.

Haven't tried the PSU yet. It's an almost new Seasonic SS-380GB. Not a cheapo.

20A @ +3.3V
20A @ +5V
2x17A @ +12V

Should be ok, right?

EDIT: I just remembered, I've measured the power draw of this system at the wall when it was my main computer. It was something like 170W at idle! My current i7 system should be around 40W idle. How times have changed.)

I'm not a power supply expert, but think I would've gone with an older power supply. Their 5V rails tend to be more powerful.

Tested my coppermine setup against my c2d and i5 systems a few weeks ago. They all idled between 50 and 70W. Load was very different tho. The coppermine maxed out at 80-90W, but the C2D and i5 went all the way up to 250-300W. I've yet to see anything that beats my Mac mini that idles at 15W. 😀

joacim wrote:

I'm not a power supply expert, but think I would've gone with an older power supply. Their 5V rails tend to be more powerful.

Yeah, don't really have one of those 🙁 I just found an old CWT 235W that supplies 22A @ 5V, that is more than the new 380W I have. Hrrrm.

Yes, Athlons (incl. early Athlon XP) & P3 systems were pre ATX12V era, and the Athlons generally ran hotter than Pentium III Coppermine & Tualatin.

Max load 20A x5V = 100W -- doesn't seem to leave much room at full CPU load. A safer rating would be 25~30A @ +5V for Athlons. Newer PSUs tend to have less on 5V and more on the 12V though, so usually you'll have to go 450W+ just to get 25A @+5V (and probably 1000W models just to get 30A 🤣). That said, you'll likely be just fine with that Seasonic, but try 'burn'/max load test to be sure.

Another concern would be the capacitors on the Abit board -- worth doing a little inspection on them (any leaking/bulging?).

And kudos to our friend oerk for still having the motherboard manual for this beast. How cool is that?

archsan wrote:

Yes, Athlons (incl. early Athlon XP) & P3 systems were pre ATX12V era, and the Athlons generally ran hotter than Pentium III Coppermine & Tualatin.

Max load 20A x5V = 100W -- doesn't seem to leave much room at full CPU load. A safer rating would be 25~30A @ +5V for Athlons. Newer PSUs tend to have less 5A and more on the 12V though, so usually you'll have to go 450W+ just to get 25A @+5V (and probably 1000W models just to get 30A 🤣). That said, you'll likely be just fine with that Seasonic, but try 'burn'/max load test to be sure.

Another concern would be the capacitors on the Abit board -- worth doing a little inspection on them (any leaking/bulging?).

There is an ISA sound card, so it might be mandatory to have an older (PIII era?) ATX PSU with -5V line intact.

And I'm not sure if you really need some high power PSU with that stuff. One of my "experimental" rigs has a PIII 1000 CPU, 2x512MB PC133 SDRAM, eight harddisks (5 SCSI+3 IDE), two floppies (3.5 + 5.25), a DVD-RAM/RW, a SCSI Backup unit, a SB AWE64+Simmconn+32MB RAM, a Matrox Millennium II with 12MB RAM, two Voodoo2 12MB in SLI, a Voodoo3 3000, an Adaptec AHA-29160 SCSI controller, a PCI 10Mbit NIC, a big AMD slotA CPU cooler, two high flow/high RPM 120mm case fans (180 cfm), and only an Aopen 350W power supply (probably produced by chepo FSP). It was (and still) working happily.

As previously mentioned, PSU wattage doesn't tell the whole story. For early Athlon boards like yours the 5V rail is most important. The later Athlon XP-supporting boards moved more load over to 12V.

I run an Athlon 1200 rig on a Seasonic SS-300FS. Even though it's only rated at 300W it has 30A on 5V and I've not had any problems.

^What bjt said.

tayyare wrote:

And I'm not sure if you really need some high power PSU with that stuff.
[...] and only an Aopen 350W power supply (probably produced by chepo FSP).

Older 300W ATX PSUs usually still give you at least 30A on +5V, which would be plenty for your system. Just check the label on your PSU.

tayyare wrote:

There is an ISA sound card, so it might be mandatory to have an older (PIII era?) ATX PSU with -5V line intact

It'll be fine as the card doesn't use -5v.
The Mystery -5V Rail

I like this forum 😀

archsan wrote:

Another concern would be the capacitors on the Abit board -- worth doing a little inspection on them (any leaking/bulging?).

Nope! That one is stable and seems to have good capacitors. I've seen A LOT of socket A boards die over the years. This one has outlived all of them. I know how to use a soldering iron though.

mwenek wrote:

And kudos to our friend oerk for still having the motherboard manual for this beast. How cool is that?

Thanks! Like I said, I bought this board new with my first paycheck 14 years ago. Crazy expensive, I think it cost around 400DM (about $250?). Had a Duron 700 @ 1049 MHz and a 60GB RAID 0 in it. Good times.

I'm trying to find a better suited PSU, though I don't think this is the culprit. The Seasonic seems to have -5V line though.

I had this board with a Tbird 700 or 750 running a little OC'd. 😀
Think I still have the manual but shouldn't have let the rig be taken away 🙁

I have 2 Kt7 boards, and all had leaky caps, change them.

Success!!! Test run after changing the PSU was completely stable! (about half an hour of UT99) Before that, I had lockups shortly after booting Windows.

JaNoZ wrote:

I have 2 Kt7 boards, and all had leaky caps, change them.

On closer inspection, you're right. The caps ARE bulging. Do you recommend changing all the electrolytics on the board?

Changing all the caps is best, unless any of them are good brands, but sadly they probably aren't. As awesome as ABit's boards were, they used terrible caps in that time period. It ruined their reputation and it's probably the reason they're not in business anymore. They switched to good caps later, but it was too late.
You can maybe skip the really small values, like 100uF and below. It depends how thorough you want to be - I'm sure there's been some cases where even smaller caps have caused problems. The larger values are the most common culprits though. Bad caps don't always bulge (and small caps pretty much can't), so given that you know you have bad caps, anything that's from a suspicious family should be replaced even if it still looks okay.

Since they're already bulging, I'd avoid using it until they can be replaced. What you have now is a known working board with dying caps, so it's totally saveable. As the caps fail, sometimes they cause collateral damage so it's best to change them before anything bad happens. Once the board completely dies, recovery after a recap is "likely", but not guaranteed anymore.

I tried taking a shortcut on this with an ABit BX133-RAID, and about 6 months later I had to finish the job when the other caps started to go. Jackcon sucks.

Gotcha, thanks for the detailed answer! I think I'm going to replace all of them, if I'm in there anyway. Sucks that I can't use it in the meantime, but that thing holds sentimental value to me so it will be worth it.

Could this be the reason the internal USB isn't working correctly anymore? Data transfers over USB fail after a few minutes.

Had an Asus K7V333(?) fail catastrophically with dying caps, and that thing was maybe 3 or 4 years old at the time. I have no intention of repeating this experience.

Is going with slightly higher values ok, if I don't have the exact same?

This is why those old school Enermax and Antec units with the strong 5v rail and -5v rail are almost required for these beastly Athlon rigs. You need the -5v rail for the ISA if I remember right and the cpu on most socket A boards pulls from the 5v rail.

oerk wrote:

Could this be the reason the internal USB isn't working correctly anymore? Data transfers over USB fail after a few minutes.

I suppose it's possible, but I don't know. Certainly bad caps don't make the chips happy, since the voltages would be fluctuating more than they're supposed to.
Early VIA USB was notorious, but since you said "anymore" it sounds like yours used to work correctly at one time.
My sister had major problems with an older (socket-7 MVP3) system using VIA USB. In her case it couldn't handle any sort of long transactions, the only thing it was good for was a mouse. Printing/scanning was dodgy, flash drives were hopeless. It sounds kind of like your problem, but that was an older system. I don't remember if she was using the 586B southbridge or an addon card to get her USB ports. It's probably not relevant - your board is probably 3 years newer so by then I'd think VIA USB would have improved.

Is going with slightly higher values ok, if I don't have the exact same?

It should be fine. The suggestions I've seen are to stay the same or slightly higher capacitance, without going crazy.
The only exception I personally ran into was on the BX133 I mentioned earlier. Initially I replaced some 1000uF caps with 1500uF, and afterward the PSU couldn't start the board without tripping overload protection. Apparently the inrush current to charge the bigger caps was too much for the PSU. That's the only time I ever had that happen though, so I might have just shorted something without realizing it. It was my first attempt at recapping and my soldering was terrible. I might also have been using some junk PSU, I can't remember.
I've upsized caps on some other boards and didn't have a problem with it. If the replacements aren't wildly bigger than the originals I'd just use them.

Usually the more important characteristic for the large capacitors is their ESR. This affects how well they filter noise out of the circuit they're attached to. They should be low ESR type capacitors, not general purpose.
It isn't necessary to dig this deep - but if you find the VRM chip on your motherboard, and find a datasheet for it, it might include a suggested circuit diagram for the CPU Vcore circuit. These suggestions often include capacitor specs, which might even lend some encouragement to changing what the board manufacturer used. To make any use of that info though, you'd have to poke around with a meter to figure out which caps match up to the input and output of that VRM.