I've installed the "new" (less old) 500W supply. I first had to splice in a second floppy power connector. Luckily the mobo has free space to accomodate a 24 pin atx connector.

All in all it was far less work and hassle than I antïcipated. Everything works.

This one is more up to the task for the upcoming Geforce 4 than the old 235W psu.

Nicolas 2000 wrote on 2025-06-11, 20:47:

I've installed the "new" (less old) 500W supply. I first had to splice in a second floppy power connector. Luckily the mobo has free space to accomodate a 24 pin atx connector.

All in all it was far less work and hassle than I antïcipated. Everything works.

This one is more up to the task for the upcoming Geforce 4 than the old 235W psu.

But is it really? I know it will work, but having a 24 pin lets me know right away, it's fairly "modern" and probably has a P4 power and maybe even a PCI-E, and those era PSUs are for the 12V computers (P4/AXP to current) and might not like the low/no load on the 12V.

The P3 is a 5V system (including the GF4). I like to see at least 25A on the +5V on a computer I care about. I usually just service an old PSU and call it good. Re-cap is pretty much all there is to do. Sometimes a fan needs to be replaced. Most of my 300W+ PSUs have 30A or MORE on 5V

I've been following, but don't remember if you've shown the PSU(s) you've been using.

The PSU on my bench right now is only 250W, but can dish-out a whopping 30A on the 5V rail (10A on 12) (1992 PSU).

I also have a "modern" unit sitting here: It's 850W, but only has 24A on the 5V (70A on 12) and the 5V is out-of-spec and squeals if the load on the 12V isn't there. In fact, if no load at all, it won't even turn on. (2013 PSU)

I've seen "modern" PSUs deriving its +5V from the +12V rail, using a circuit like this:

The PSU tends to be overlooked, yet I believe it to be the most important part of the computer. Using a modern PSU should help keep you away from component failure and probably have better filtering.

It's been fun watching your progress. Good work! Its really come together nicely. Do you name your computers? I know Windows probably did, 🤣 It says T-Bird on the front, but no t-bird.

Major Jackyl wrote on 2025-06-11, 23:09:

But is it really? I know it will work, but having a 24 pin lets me know right away, it's fairly "modern" and probably has a P4 p […]
Show full quote

But is it really? I know it will work, but having a 24 pin lets me know right away, it's fairly "modern" and probably has a P4 power and maybe even a PCI-E, and those era PSUs are for the 12V computers (P4/AXP to current) and might not like the low/no load on the 12V.

The P3 is a 5V system (including the GF4). I like to see at least 25A on the +5V on a computer I care about. I usually just service an old PSU and call it good. Re-cap is pretty much all there is to do. Sometimes a fan needs to be replaced. Most of my 300W+ PSUs have 30A or MORE on 5V

I've been following, but don't remember if you've shown the PSU(s) you've been using.

The PSU on my bench right now is only 250W, but can dish-out a whopping 30A on the 5V rail (10A on 12) (1992 PSU).

I also have a "modern" unit sitting here: It's 850W, but only has 24A on the 5V (70A on 12) and the 5V is out-of-spec and squeals if the load on the 12V isn't there. In fact, if no load at all, it won't even turn on. (2013 PSU)

I've seen "modern" PSUs deriving its +5V from the +12V rail, using a circuit like this:

The attachment 20250611_172135.jpg is no longer available

The PSU tends to be overlooked, yet I believe it to be the most important part of the computer. Using a modern PSU should help keep you away from component failure and probably have better filtering.

It's been fun watching your progress. Good work! Its really come together nicely. Do you name your computers? I know Windows probably did, 🤣 It says T-Bird on the front, but no t-bird.

Actually rated power or 5v current means very little in case of using modern PSU to run old system. What matters is its type.

If you try to use group-regulated power supply it will have issues. Even if it is rated for stupid amount of power and very decent 5V current - once you load it with 5V and near-zero 12V load voltages will drift (5V down, 12V up), possibly to a point they may go out of spec, cause stability issues or cause PSU to shut down. This are unsuitable for 5V system, regardless of ratings. It is not about the total load here, but distribution of load between 5V/12V.

If you use DC-DC supply, the one using those module you've shown, it will work absolutely great. Yeah, those module will have to work harder than usual - but as long as it is withing spec in terms of max current the voltages will stay where they are and output it produces will very likely be much cleaner/better than that of period-correct PSU.

The best thing - they tend to use ~similar converters in all PSUs - so no reason to buy really powerful one, 400-500W from good brand will do. I have 400W BeQuiet PSU like this running old socketA system with 5v VRM for years and it works great - voltages are stable and it does not even spin up the fan too much. And socketA is the worst case as it is quite power hungry, P3 system should cause no issues at all to pretty much any DC-DC PSU.

I've compared both psu.

OLD - NEW

5V 22A 20A (there we go)
5VSB 0.8A 2.5A
3.3V 14A 24A

5V+3.3V combined unknown vs 120W

12V 8A 2x20A

So yes, I do indeed loose a tiny bit on the 5V rail with the new PSU.

There is a (tiny) bit of 12V load: 4 case fans constantly at max, the Audigy front bay, 5400rpm sata drive, 2 dvd and a floppy drive.

I don't know how the Leadtek geforce 4 ti4200 agp will load the 3.3V, 5V and 12V compared to the rage128pro. I think 3.3V is an important one for AGP?

Do measure 12v/5v with a multimeter, just to be safe. And do keep in mind that for such sustem with CPU/GPU using 5v this is effectively 120W power supply at best.

Just a voltage measurement in a free molex when running?

Yes, that should be enough.

Regarding the name of the computer, in Windows it is a descriptive name that is outdated already.

In real life, it's the Beige Baller Build. And for that reason alone, I'll try to print some of those square case front stickers for P3, Voodoo2 etc. Like we did back in the day. 😀

Nicolas 2000 wrote on 2025-06-12, 07:22:

In real life, it's the Beige Baller Build. And for that reason alone, I'll try to print some of those square case front stickers for P3, Voodoo2 etc. Like we did back in the day. 😀

You can get some pretty faithful replicas of those stickers on eBay.

Just search for "pentium sticker" and pick the one you like. I think there are some for Voodoo cards as well.

They are indeed, but quite pricey. I used to print them myself back in the day as well. 😀

I've unplugged the CDRW to free up a molex for measurements. I've measured while running 3DMark99, both the rage and the Voodoos.

I measure using a high end Brymen DMM which provides a DC measurement with an AC component (ripple). Slow changes in the DC measurement are not shown in the ripple measurement, which is for faster changes.

12V: 12.39-12.48V, ripple 0.00V (after measuring for a while)
5V: 4.85-4.9V, ripple 0.004V (Rage)/0.018V (Voodoos)

So both rails are within the 5% ATX spec. Ripple is low. Seems good to me. This is with the 733CPU and not yet with the Geforce.

That should be fine, however you can see that 5V is starting to drop, 12V starting to rise. Still within spec, but add more load on 5V and it may become an issue. Probably should check it again with faster CPU and the videocard...

I will, it's easy enough to check now that the case is still open. Measuring while running 3dmark99 should be about as relevant as it gets; it's not reading from cd at that moment but apart from that it's all hands on deck.

The Geforce arrived. What do we think about the green capacitor?

That's dead and needs replacement.

How dead? Well, anywhere from completely to just being significantly out of spec.

Also look carefully around those other caps for signs of any residue on the PCB - those can leak instead of bulging.

I'll have to order replacement cap(s), all I have is way larger voltage spec so physically too large.

Safe to test the card first as-is to know whether it works at all before I let loose on it with the soldering iron, or better to keep it asleep until that cap has been replaced?

Nicolas 2000 wrote on 2025-06-13, 10:36:

The Geforce arrived. What do we think about the green capacitor? […]
Show full quote

The Geforce arrived. What do we think about the green capacitor?

GSC are well known bad caps... I had to pull a load off my FX5200 to make it work. It artefacted like mad (unicorn puke) until I recapped it. Now it's perfect. I'd replace any GSC caps on the card.

Testing with known dead caps always creates a small risk of killing something. I usually try not to do it if i value the part, faulty caps are obvious and are not too much work to replace. Sometimes i do test things with dead caps though, mostly when the condition is totally unknown and do not feel like replacing a lot of capacitors, like whole motherboard worth, for nothing.

In this case i'd just swap the one dead cap, it's fast and easy enough...

Ok that means this card will have to wait a few weeks until I have enough to order. Shipping costs for 1 cap...

Could always salvage a cap from somewhere. Dead motherboard, dead PSU, something. I like to buy piles of modern-ish dead stuff locally from time to time for this reason. Anything from caps to some chips, connectors, whatever can be salvaged and it saves a lot of money and time.