nd22 wrote on 2025-02-06, 09:54:

Athlon 1400C + 2*1gb DDR + gf3 ti 500 + Abit KG7-raid + HDD + DVD = 30A required on the 5V rail + 180W combined on the 3.3V+5V rail.
Corsair RM1000i : 4.95V measured on the 5V rail while running 3dmark 2001 in a loop!
Cons to this power supply: 200 euros new!!

Yeah, but at that point, might as well just hunt for a good NOS PSU from back in the day: https://www.anandtech.com/show/1124/1
If you're patient & lucky, you can still find these new, sealed, for less than $30, which is definitely worth it.

bloodem wrote on 2025-02-06, 10:08:

nd22 wrote on 2025-02-06, 09:54:

Athlon 1400C + 2*1gb DDR + gf3 ti 500 + Abit KG7-raid + HDD + DVD = 30A required on the 5V rail + 180W combined on the 3.3V+5V rail.
Corsair RM1000i : 4.95V measured on the 5V rail while running 3dmark 2001 in a loop!
Cons to this power supply: 200 euros new!!

Yeah, but at that point, might as well just hunt for a good NOS PSU from back in the day: https://www.anandtech.com/show/1124/1
If you're patient & lucky, you can still find these new, sealed, for less than $30, which is definitely worth it.

This is how i got the Enermax Eg701AX PSU: 34A on the 5V rail, enough for any socket 462 system!

bloodem wrote on 2025-02-06, 10:08:

Yeah, but at that point, might as well just hunt for a good NOS PSU from back in the day: https://www.anandtech.com/show/1124/1
If you're patient & lucky, you can still find these new, sealed, for less than $30, which is definitely worth it.

NOS PSU that is something like two decades or more old isn't reliable, capacitors go bad unused too. I have one vintage NOS PSU incoming and I won't use that on any of my systems until I've recapped it.

On that note, I only use vintage PSUs with these 5V heavy setups as they are still easily available and are cheap, I just recap the secondary side by default. I find using modern beefy PSUs a waste with these systems and even if you have nominally enough power on 5V and 3.3V rails, group regulated PSUs may still struggle when there is little load on 12V as they were designed to provide most power on 12V instead of 5V/3.3V.

CharlieFoxtrot wrote on 2025-02-06, 10:42:

NOS PSU that is something like two decades or more old isn't reliable, capacitors go bad unused too. I have one vintage NOS PSU incoming and I won't use that on any of my systems until I've recapped it.

Capacitors don't go bad when unused (this is mostly a myth), unless they are cheap Chinese caps from the capacitor plague era.
If you have a good quality tester, then it's quite easy to check the health of existing capacitors, even in circuit.
Anyway, the NOS PSUs that I got so far have not had a single issue with caps, but they all came with high quality Japanese caps from the factory, which were less prone to the capacitor plague issues (though some Japanese companies were still affected).

CharlieFoxtrot wrote on 2025-02-06, 10:42:

I find using modern beefy PSUs a waste with these systems and even if you have nominally enough power on 5V and 3.3V rails, group regulated PSUs may still struggle when there is little load on 12V as they were designed to provide most power on 12V instead of 5V/3.3V.

Indeed, definitely agree with this.

bloodem wrote on 2025-02-06, 10:59:

Capacitors don't go bad when unused (this is mostly a myth), unless they are cheap Chinese caps from the capacitor plague era.

They absolutely do and that is a fact. All electorlytic caps go bad with time. They slowly dry out decreasing the capacitance and increase ESR. In a PSU, this at first will cause higher ripple which alone is a bad thing. Shelf life of a cap depends a lot from storing conditions, but large reputable sellers such as Digikey and Mouser generally have around 2 years shelf life for the capacitors. Caps for sure last much longer than that, but they make sure that they sell caps that will still last years in operation after leaving the shelves as age alone kills them slowly.

As far as the origin of the caps in these PSUs, most of the main stream manufacturers in early 2000s used something other than Japanese caps. CEC and Teapo are very common in different brands (they aren't disasters, but still something that can't be considered "high quality"). PSU manufacturers also used caps from different manufactures through the years. For example, at one point Antec used Fuhjyyu which turned out to be a complete dumpster fire until they went back to something more reliable. Anyways, if you happen to bump into a NOS PSU, you need to do quite a bit research what caps that particular PSU model possibly has inside. The time you spend on the research is probably more than you spend on the recap process itself.

If you have a good quality tester, then it's quite easy to check the health of existing capacitors, even in circuit.

With correct type of LCR meter, yes, you can normally find out if the capacitor is already in very bad state in circuit although reliable numbers are still guranteed out of circuit. It doesn't still remove the fact that the caps have finite life. If you have the caps at hand, it is 0,5-1 hour job to recap a psu and you are good to go for a long time in the future. It is so simple job that I haven't found a reason not to do that as a default and as a preventive maintenance. And not all people have those meters and with a price of one decent LCR meter you can recap several PSUs with quality caps again ensuring proper operation far into the future.

Anyway, the NOS PSUs that I got so far have not had a single issue with caps, but they all came with high quality Japanese caps from the factory, which were less prone to the capacitor plague issues (though some Japanese companies were still affected)

The fact that you haven't had issues with them so far doesn't mean that 20 years old caps should be trusted and caps don't go bad sitting on a shelf. You also don't know how that PSU has been stored along the years and conditions are a huge factor how long those caps will stay in spec. Have they been in a cold storage? In a hot and humid environment? Who knows. I personally mainly have used second hand Enermax PSUs and I've mesured many of the old caps after removing them and they have been okay. I still have zero trust that they will still last good number of years. I didn't claim that old caps are certainly bad after 20 years, only that no one should trust them because they are unused.

But hey, everyone can plug whatever they want to their hardware, even the crappiest and cheapest old PSUs, so I'm completely fine with either way. I just decided to chime in that no one should trust these old PSUs, used or not, and they will certainly age like pretty much any electornics just when sitting.

Well I cound't myself lucky browsing ebay a few years ago I snagged a bunch of Antec True Power 350 and 400 units with good caps that check out, they are what i use now in my builds

CharlieFoxtrot wrote on 2025-02-06, 20:49:

They absolutely do and that is a fact. All electorlytic caps go bad with time. They slowly dry out decreasing the capacitance and increase ESR. In a PSU, this at first will cause higher ripple which alone is a bad thing. Shelf life of a cap depends a lot from storing conditions, but large reputable sellers such as Digikey and Mouser generally have around 2 years shelf life for the capacitors. Caps for sure last much longer than that, but they make sure that they sell caps that will still last years in operation after leaving the shelves as age alone kills them slowly.

Everything will go bad, given enough time. Even black holes will eventually radiate their last particle. 😀

Anyone with extensive experience in electronics repair will likely agree with me that faulty capacitors are the exception rather than the norm - even in 50+ year old devices. Unfortunately, the capacitor plague instilled a widespread misconception that capacitors have inherently short lifespans, whereas, before that time, capacitor failure was rarely a primary concern.

I've lost count of how many times I've preemptively replaced capacitors in 35+ year old cheapo Chinese AT power supplies (see attached pics), only to test the old ones and find them still within spec. Funnily enough, in some cases, the original capacitor even had a lower ESR & voltage drop than the brand-new Nichicon or Panasonic replacement.
My point is that, even if you should always make sure that caps are still in good shape, if you do find a 20 year old sealed PSU from a known good brand (one that hopefully did not use capacitors affected by the capacitor plague), you most likely will be able to use it for many years to come without any fear of failing caps.

As a rule of thumb (based on Arrhenius' Law), a capacitor’s lifespan doubles for every 10C degree reduction in operating temperature. So, if you have a good capacitor that's rated to run for 2000 hours @ 105C, you are looking at 500k hours/~60 years at room temperature (and that's with actual usage). There's also the concept of reforming, where a capacitor that may have drifted out of spec after decades of storage has a real chance of restoring its oxide layer when voltage is reapplied (even if not doing it in a controlled & gradual manner, as reforming should normally be done).

nd22 wrote on 2025-02-06, 09:54:

Athlon 1400C + 2*1gb DDR + gf3 ti 500 + Abit KG7-raid + HDD + DVD = 30A required on the 5V rail + 180W combined on the 3.3V+5V rail.

Hmmmm???? How are you getting those numbers?
That Athlon CPU is 72W TDP. The GF3 TI500, judging by the heatsink size is 15 Watts... maybe 20 Watts tops. So that gives a total of less than 100 Watts... but let's round it up to 100 for fun. Assuming all of this comes from the 5V rail (though I'm not 100% sure if the TI500 actually pulled its GPU power from 3.3V or 5V rail, but whatever, let's ignore this little detail), that's still 20 Amps at the most, and with a bit of room to spare. An HDD + DVD ain't gonna add up another 10 Amps. So where you get the 30 Amp requirement, I'm not sure. If it's from some online calculator, take it with a spoonful of salt... maybe even an entire glass.

I always do my own calculations based on the CPU and GPU TDP (and sometimes mobo chipset TDP, if it's a more power-hungry one.) Also, unless I go total overboard with the HDDs and ODDs (and only if I plan to use them all at the same time) do I consider adding all of their power requirements to the total. For the most part, 1 HDD will draw about 500 mA from the 5V rail and about just as much from the 12V rail when up and running. During spin-up, the 12V rail current can be high (up to 2 Amps), but that isn't of too much concern with one or two HDDs.

BTW, I also NEVER trust the label on the "no-name" PSU manufacturers. Instead, I open them and check the part numbers of all of the rectifiers. This gives me a much better idea of what the PSU is capable of on each rail. A lot of times, the cheapo PSUs will write sky-high numbers for the 3.3V and 5V rails, which is not true at all. Take the L&C B300ATX PSU mentioned in the beginning of this thread. I just got one like it last week in a retro PC I picked up for next to nothing. Both the 3.3V and 5V rail promise close to 30 Amps each. Yet, get this: the rectifiers in this PSU for the 3.3V and 5V rail are only 10 Amp and 16 Amp parts! 🤣 So the label is a complete lie.

In regards to these Segotep PSUs everyone has been talking about here in the last few pages - I wouldn't trust the label on those either. FWIW, I bought a used PC for my nephew last summer, and it came with one of these Segotep PSUs - model GTR550, for anyone that cares. Of course I opened it for inspection (and cleaning) and I have to say, I wasn't that impressed... though I can't say it's a terrible PSU either. Basically, it was exactly what was popular 25 years ago - an ancient half-bridge group-regulated design and probably the same mediocre efficiency too. Only few things that distinguished it from the cheapo PSUs back in the day: it had complete input and output filtering... though cap-wise, it was full of "no-name" cap brands I didn't recognize, so I doubt it will last too long. Apart from that, it wasn't built too badly - 13009 NPN BJTs in TO-220 case (so probably going to act like "better made" 13007's) and decent output rectifier setup for the 12V rail (probably to cater to modern 12V-based systems). The rectifiers for the 3.3V and 5V rail were more or less rated for what the label promised, but I wouldn't suggest anyone to push it that high on those rails... which going back to the topic at hand here (PSUs with strong 5V rails), I wouldn't say I recommend this Segotep PSU for a 5V-heavy setup.

Nemo1985 wrote on 2025-02-01, 23:01:

Unlucky I tried the tagan and it did a bang probably it didn't like being in a basement for 10 years...

Probably needed a recapping.

From what I remember, those higher-end Tagans were actually pretty good PSUs, but just riddled with crap caps... which sitting for 10 years probably got even worse over time. Likely the 5VSB circuit blew. If it was mine, I would have opened it before powering it on (it's actually a rule for anything I get used - everything must be opened and inspected first before being plugged in.)

CharlieFoxtrot wrote on 2025-02-06, 10:42:

bloodem wrote on 2025-02-06, 10:08:

Yeah, but at that point, might as well just hunt for a good NOS PSU from back in the day: https://www.anandtech.com/show/1124/1
If you're patient & lucky, you can still find these new, sealed, for less than $30, which is definitely worth it.

NOS PSU that is something like two decades or more old isn't reliable, capacitors go bad unused too. I have one vintage NOS PSU incoming and I won't use that on any of my systems until I've recapped it.

On that note, I only use vintage PSUs with these 5V heavy setups as they are still easily available and are cheap, I just recap the secondary side by default. I find using modern beefy PSUs a waste with these systems and even if you have nominally enough power on 5V and 3.3V rails, group regulated PSUs may still struggle when there is little load on 12V as they were designed to provide most power on 12V instead of 5V/3.3V.

+1

BTW, only time a NOS PSU might be OK to use as-is is if came with Japanese capacitors originally. Some of the Dell and HP OEM PSU from the Pentium II and 3 era were built like that. Only problem is, some of these were proprietary.

bloodem wrote on 2025-02-07, 07:04:

Anyone with extensive experience in electronics repair will likely agree with me that faulty capacitors are the exception rather than the norm - even in 50+ year old devices.

I'd say that depends a lot on the specific device we are talking about.
Old Japanese caps tend to have pretty good record for the most part and indeed don't age too badly at all. For example, I was servicing my dad's Sony stereo amplifier from the late 70's, and most of the large caps were well within spec, despite its age (the amp was used regularly all these years, though... so it wasn't sitting without use.) At the same time, I've ran into quite a few Nichicon PR series with electrolyte leaking from the bungs (known defect with age for this series.)
That said, most non-OEM PSUs from the 90's and onwards rarely used Japanese caps. Instead, they used "2nd tier" Taiwanese caps like Teapo, CapXon, Ltec, and etc... which indeed weren't all too terrible... but then some really were. The problem with many of the "2nd tier" caps is that you never know whether a particular batch will last or not. And then there's the application too. For example, the CWT-built Antec PSUs with Fuhjyyu capacitors used to fail A TON... yet, I've ran into other PSU manufacturers that used Fuhjyyu and didn't really have any problems. So clearly the design of the PSU matters too.

But either way, I think we can all agree here that a 20 YO... or even a 15 YO NOS PSU shouldn't be trusted blindly and just put into a system without any regards.
At the same time, buying a new PSU that's not really designed for a heavy 5V-load is also not a great choice. This is why I too support the idea of getting older PSU and encourage people to recap them. It's not too hard of a process (especially after you do a few... save for some of the really packed "double-deckers" 🤣 - those are always slow to work on).

bloodem wrote on 2025-02-07, 07:04:

Unfortunately, the capacitor plague instilled a widespread misconception that capacitors have inherently short lifespans, whereas, before that time, capacitor failure was rarely a primary concern.

That's indeed true to an extent. I've seen a lot of people worry or toss their cap stock in the bin when it's sat for more than 5 years... which is a real shame, since most capacitors don't age that badly sitting on the shelf, unless they were garbage quality to begin with. At this point, I have some caps in my stock that I got 15+ years ago from Digikey/Mouser. So far, all of them have been OK, despite being on the shelf for so long. I do reform them before installing them in a device, of course - just in case. I think I didn't reform them only a few times, where I put them in devices for my own use where I didn't really care that much and just YOLO-repaired them. 😁

momaka wrote on 2025-02-08, 00:15:

Hmmmm???? How are you getting those numbers? That Athlon CPU is 72W TDP. The GF3 TI500, judging by the heatsink size is 15 Watts […]
Show full quote

nd22 wrote on 2025-02-06, 09:54:

Athlon 1400C + 2*1gb DDR + gf3 ti 500 + Abit KG7-raid + HDD + DVD = 30A required on the 5V rail + 180W combined on the 3.3V+5V rail.

Hmmmm???? How are you getting those numbers?
That Athlon CPU is 72W TDP. The GF3 TI500, judging by the heatsink size is 15 Watts... maybe 20 Watts tops. So that gives a total of less than 100 Watts... but let's round it up to 100 for fun. Assuming all of this comes from the 5V rail (though I'm not 100% sure if the TI500 actually pulled its GPU power from 3.3V or 5V rail, but whatever, let's ignore this little detail), that's still 20 Amps at the most, and with a bit of room to spare. An HDD + DVD ain't gonna add up another 10 Amps. So where you get the 30 Amp requirement, I'm not sure. If it's from some online calculator, take it with a spoonful of salt... maybe even an entire glass.

I always do my own calculations based on the CPU and GPU TDP (and sometimes mobo chipset TDP, if it's a more power-hungry one.) Also, unless I go total overboard with the HDDs and ODDs (and only if I plan to use them all at the same time) do I consider adding all of their power requirements to the total. For the most part, 1 HDD will draw about 500 mA from the 5V rail and about just as much from the 12V rail when up and running. During spin-up, the 12V rail current can be high (up to 2 Amps), but that isn't of too much concern with one or two HDDs.

Let's see:
1. https://www.youtube.com/watch?v=efK7mw8eYiE Phil demonstrated in this video with the exact same setup - Athlon 1400C + geforce3 ti 500 how much the 5V rail is important in a socket 462 system and that indeed it is required some 30A on the 5V rail.
2. https://outervision.com/power-supply-calculator the best PSU calculator and I think the only one available that will show you how much A you need in each line.

nd22 wrote on 2025-02-10, 07:24:

Let's see:
1. https://www.youtube.com/watch?v=efK7mw8eYiE Phil demonstrated in this video with the exact same setup - Athlon 1400C + geforce3 ti 500 how much the 5V rail is important in a socket 462 system and that indeed it is required some 30A on the 5V rail.

Good video, but... NO, it's WRONG to jump to the conclusion that you absolutely need a PSU with a 30A 5V rail for a socket A rig.
I've done these same tests as Phil with my Kill-A-Watt meter (measuring power draw from the wall) for many of my (now retro) systems for many many years (at least as far back as 2014.)

My two most 5V-power hungry rigs right now are my Athlon XP 2500+ with a Radeon 9700 (which, BTW, is A LOT more 5V and 3.3V -heavy than the GF3 TI500 - about 20-25W more in TDP) and a 1.7 GHz Pentium 4 Willamate (72W TDP, IIRC) on a ECS P4VXASD2+ (no 4-pin 12V CPU connector) with a GF FX5600 (heavy draw from the 3.3V rail.)

Both of these I've ran for years on a Morex 230 Watt PSU from 1999 that has a 5V rating of "only" 22 Amps. Of course, I did recap this PSU, because it otherwise would have not worked at all (was a freebie pickup from Craigslist, with a seized fan and bulged caps, so it was obviously not working when I got it.) After that, it's been one of my 5V-heavy reliable workhorses. But here's an interesting point to consider: the P4VXASD2+ system came with a gutless wonder "420 Watt" TurboLink PSU that technically could do only 15 Amps on the 5V rail (as that's what the 5V rail rectifier was rated for.) Even worse, the filtering on this PSU's 5V rail consisted of only 2x 470 uF caps, which is abysmally small. Yet, judging by the amount of dust in the PSU, it looks like the PC ran for quite a while for the original owner. Of course, I'm pretty sure it wasn't a bug-free experience and probably the reason I found this PC next to a dumpster - it likely crashed frequently. I don't know what GPU and how many HDDs the original owner had in there as these were gone from the system... but I can imagine that even with the lightest power-drawing components, the 5V rail on that PSU was likely running on the edge of shorting the 5V rectifier. But, the point is, the system still ran OK for the most part. Once in my hands, obviously there was no way I'd run it with this POS PSU other than to see if the mobo POSTed (it did). When I confirmed the mobo worked, I swapped the PSU with a 250 Watt Dell HiPro 250W PSU, and that was the end of that... well not really - I also upgraded the TurboLink's 5V rectifier to a 20 Amp one and installed 2x 2200 uF caps on the 5V rail (as well as added/upgraded the capacity on the other rails too.) This brought this otherwise gutless PSU to a much more usable condition... and in fact, I tried it on a similar 5V-heavy system, and all was fine with it now.

BTW, going back to Phil's video above... I find it a little ironic that I have exactly the same Macron Power MPT-401 PSU as Phil. I also have its smaller cousins - the MPT-301 (two of them). These are indeed nice 5V-heavy-capable PSUs, but with one BIG caveat: THEY ABSOLUTELY NEED RECAPPING (putting this in all-caps for anyone that has one of these and is thinking of using it.) Macron used some of the worst crap cap brands in their PSUs (Fuhjyyu, GL, CS, and a few more brands that escape my head now), so these PSUs are guaranteed to fail 100%. I do hope Phil has recapped his. Otherwise it's a time-bomb waiting to take out his retro hardware. I actually have an interesting story for my 2nd Macron MPT-301: it was a "for parts or not working" $1 ebay special. Long story short(er), the seller of that PSU tried to use it for a retro socket A build (Duron 900)... and it all ended up in a huge cloud of smoke. So he put the PSU on ebay for parts for $1, stating what had happened. No one bid on it, except for me (a PSU-rebuild-ahollic, if there ever is such a hobby.) What's crazy is that he put it with "free shipping", so he paid out of pocket to ship it to me. Once I got it and fixed it, I then asked him if he wanted it back for a retro system again. But at that point, he said he had already moved onto building a different retro system. So I asked him if he still had the original motherboard, as I was curious to see if I could fix it as a challenge... and to my surprise, he said he did. He shipped that to me, again, out-of-pocket... so as a gratitude, I asked him if he'd be interested in a free s478 P4 board with a 2 GHz P4 that I had at the time. He did, so I sent him that and got his broken socket A board. From this, I was able to conclude that the out-of-spec ripple from the PSU was so bad that it not only took out the CPU (shorted core), but also fried (shorted) a RAM module *and* the power contacts in the RAM slot.

So morale of the above story: DON'T ignore bad caps and DO recap your oldschool PSU's, just in case... or at least if they are know to use really terrible cap brands. And #2... again, NO, you don't absolutely need 30 Amps on the 5V rail for a socket A system. Even 25 Amps is more than necessary. What's needed is an older PSU that is built to handle a majority of its load from the 5V rail *and* to make sure that this old PSU is OK internally (i.e. checked for bad caps, seized fan, heavy dust buildup, and etc.)

As for modern PSUs and why they often don't do well with a heavy 5V or 3.3V load - that's because most of them are either geared to supply most of their current from the 12V rail (often the case of group-regulated designs, where the 12V rail can go too high and the 5V rail too low with a 5V-heavy load) *or* they use DC-DC step-down buck converters for the 5V and 3.3V rails (powered from the 12V rail) which have a pretty hard "cut-off" point once you approach their load limits - i.e. their ripple tends to shoot up quickly towards their load limits. And this is very likely why Phil (and maybe even some of you here in this thread) might run into such an issue with using a new PSU to power an old 5V-heavy power rig.

Another point I'd like to mention about Phil's video above, though probably much less important, is that the test motherboard he was using appeared to use the stock/original OST capacitors. These aren't totally terrible, but I have seen them go bad before and do know they can be a source of problem too. So if any were failed or marginal, that too could have contributed to why he was not able to get the board to POST/work with some of the higher-power CPUs & GPUs. In the case of my P4VXASD2+ mobo, its caps were still all good and well within spec... which probably made up quite a bit for the otherwise out-of-spec ripple that the original TurboLink PSU was spitting out at it.

this is what I use with my socket 462 build, it's a modern PSU from Lenovo.

Only problem is that I need to use Floppy and Molex adapters connected to the SATA connectors on the PSU, but these are cheap to buy.

I do not know where to put this so here it goes:
Today, 22 of February 2025, the only power supply calculator that showed exactly how much do you need for your socket A/462 machine or for your LGA 775 system stopped working and was taken offline 🙁.
https://outervision.com is no more!
I saved a lot of possible configurations beforehand so I hope I can post them up here if anyone will ever need some data for his system.
Wayback machine is broken and can't load the calculator to actually show you how much Amperes do you need on the 5V rail or the overall wattage required for your system.