MattRocks wrote on 2026-04-11, 11:57:

Luckily, I have an original ATX PSU from the AT/ATX transition era. I was putting that ATX PSU into an AT case by simply remapping the reset button to ATX power button. But, I'm glitched - the world is upside down! The ATX PSU doesn't actually fit the AT case because the inside intake vents that should face-off to the CPU are facing-off to the top of the case.

I haven't heard about this before, but now I'm surprised too.

Luckily, there are still plenty of ATX PSUs with 80mm fan on the bargain.

If you need ATX PSU with 120mm fan and an inverted mounts, you'll probably find it in branded computers (IBM/Lenovo, HP, etc.). There's a similar confusion there - the PSU mounts is upside down.

ott wrote on 2026-04-11, 15:09:

Luckily, there are still plenty of ATX PSUs with 80mm fan on the bargain.

If they need servicing, they aren't a bargain - I had one blow up with a bang.

In either case, mine is an 80mm ATX PSU. Photo attached, showing the ATX assumption is the inverse of AT reality.

MattRocks wrote on 2026-04-11, 16:06:

ott wrote on 2026-04-11, 15:09:

Luckily, there are still plenty of ATX PSUs with 80mm fan on the bargain.

If they need servicing, they aren't a bargain - I had one blow up with a bang.

In either case, mine is an 80mm ATX PSU. Photo attached, showing the ATX assumption is the inverse of AT reality.

Plenty of room for that fan to breathe tho.
And don't forget that in early ATX the PSU-fans were supposed to do double duty, both cooling the PSU and the CPU.
The AT "standard" is basically the opposite of ATX, AT cpu's were (usually) placed at the air intake of the case while ATX had it placed near the PSU to allow OEM's to cheap out on fans =)

Intel Socket 7 (AT era) seems to consistently position the CPU above the PCI slots, which is closer to Intel ATX thinking - only place for the heat to go is up.

AT's only free space for an extra fan is the front intake and the CPU isn't near that. It's all rather inelegant.

In my opinion, the power supply should be set up the other way around—the vents in the casing should draw hot air in from the case, and the fan should blow it out of the power supply. In the photo, it’s the other way around, unless you also have vents on the bottom.

Ydee wrote on 2026-04-12, 09:14:

In my opinion, the power supply should be set up the other way around—the vents in the casing should draw hot air in from the case, and the fan should blow it out of the power supply. In the photo, it’s the other way around, unless you also have vents on the bottom.

You are saying the same thing as me.

However, an ATX PSU can't go the other way up without cutting the rear of the AT case to make space for the socket and fan swapping sides - note the PSU opening in the AT case has a rounded side for fan, and square side for power sockets! Now imagine flipping the PSU over - doesn't fit, right?

In 1995, Intel wrote Advanced Technology eXtended v1.0 upon which everything should change, but most of the changes are just different for the sake of being different. We are now on ATX v3.x, and there is still no centralised archive or library with historical specs.

I doubt many people have been tracking the changes - or realised Intel keeps setting new ticking time bombs to contain PC life expectancy.

Some older ATX power supplies don't have air intake vents on the bottom, but rather on the side opposite the fan and the power cable socket. These might be compatible with your case. Like this one:
https://computercyber.com/products/new-origin … b-atx-ps-7270c/

Ydee wrote on 2026-04-12, 13:01:

Some older ATX power supplies don't have air intake vents on the bottom, but rather on the side opposite the fan and the power cable socket. These might be compatible with your case. Like this one:
https://computercyber.com/products/new-origin … b-atx-ps-7270c/

That would mount correctly and it's a good suggestion for the problem as I described it, but I don't think it suits my world.

The EVOC PS-7270F actually an industrial server PSU.

Part of the reason I can embrace swapping AT original for an ATX PSU on my AT/ATX transition i430TX is that I see value in audio capabilities. The i430TX isn't strong for audio production, but it's solid for playback - I might install my Dxr2 kit and a sound card that natively supports 44KHz CD + 48KHz DVD sources. That is where an industrial server PSU might clash.

Home AT machines had a passthrough power cable so one whole complete PC system was conveniently powered from exactly one plug - even AT era active speakers had passthrough splitter cables. The EVOC breaks that feel. Home ATX machines break that feel too, but introduce a different home feel - more heat, bigger fans, quieter cooling.

I'll give it some thought..

Of course, that link was just for illustration—to show a type of PSU with holes on the sides rather than on the bottom. Depending on your needs, you can also find other high-quality PSUs that use this type of cooling system.

MattRocks wrote on 2026-04-11, 11:57:

And given the PSUs have changed, and many flavours are no longer made, should old PSUs now be given more TLC? Should I even be repairing my AT PSU?

If you have the time and the skills (or are willing to learn), why not?
Otherwise, you can just chuck them back on the bay or whatever other site you prefer and hope someone else who's into PSU/old PC restoration to find them and buy them.

I personally prefer to restore PSUs almost all of the time - even the not-so-great ones. Only the really crappy gutless wonders with poor designs and PCB layout I leave for parts. Just about everything else gets fixed (but slowly 🤣).

And by the way, that PSU on the picture in your 2nd post looks like an FSP (Fortron/Source/FSP Group) and a rather old unit at that, judging by the heatsinks seen through the vents and the stickers on top ("W/ Noise Killer"). No surprise if it exploded or didn't work after all this time. These older FPSs often used Fuhjyyu -branded capacitors inside that always failed. The 5VSB in particular is known to commit suicide when its output caps go bad. A lot like the FSP in this thread:
(SOLVED) FSP300-60GTF repair - blown transoptor, resistors, worth it?

That said, the nice thing about some of these older ATX PSUs is that they often use widely available parts that are still manufactured today, which makes sourcing the parts for the repair a lot easier to find.

momaka wrote on 2026-04-13, 11:56:

... by the way, that PSU on the picture in your 2nd post looks like an FSP (Fortron/Source/FSP Group) and a rather old unit at that, judging by the heatsinks seen through the vents and the stickers on top ("W/ Noise Killer"). No surprise if it exploded or didn't work after all this time. These older FPSs often used Fuhjyyu -branded capacitors inside that always failed. The 5VSB in particular is known to commit suicide when its output caps go bad. A lot like the FSP in this thread:
(SOLVED) FSP300-60GTF repair - blown transoptor, resistors, worth it?

That said, the nice thing about some of these older ATX PSUs is that they often use widely available parts that are still manufactured today, which makes sourcing the parts for the repair a lot easier to find.

Thank you for sharing.

The one that blew up was a Mercury, which is one of the worst PSUs. I suspect Mercury may have been a high street store's own label for whatever was cheapest in the grey market. The one in the photo might be an FSB (would that be written on the PCB?), but is actually branded AOpen on the outside and came in an AOpen AH series case.

What I have done is order my PSUs by cheapness, and then pair the more premium PSUs to CPUs matched by era:

AT/ATX transition MMX <-> 5V-focussed AOpen FSP300-60GT
AMD K6-2 ATX <-> 5v-focussed Delta DPS-400WB C
Athlon XP 2400+ <-> 5V/12V transition Antec TruePower Trio TP3-550
Athlon 64 2400+ <-> 12V-focussed Corsair RM750x

Having read the thread linked above, I feel somewhat... unsure... because I'm not an electrician.

Attached is a photo so you can see it has a fair number of expansion cards that I don't want to smoke. If the AOpen FPS PSU fails, is there any risk of damage to the components in my AT/ATX transition tower or is protecting?

If you get a high enough wattage ATX PSU, the 5v rails will generally be strong enough.

I've also had no issues thus-far using ATX power supplies with ATX to AT adapters whenever possible.

The only issue is if you need -5v. I used to buy the ATX to AT and ATX to ATX cables that has the -5v circuitry added in-line but the seller stopped making them a few years ago.

Now you have things like the "voltage blaster" which takes up an ISA slot to add the -5v.

You can also just add the -5v to your PSU inline, and there are instructions out there.

cyclone3d wrote on 2026-04-16, 21:20:

If you get a high enough wattage ATX PSU, the 5v rails will generally be strong enough.

Good to know. It would probably boot, but it's also a bit weird because you'd only use a small part of the new PSU.

The Corsair 750W PSU (compliant with standard ATX12V 3.1) delivers max 150W for 3.3v and 5v combined, so that's optimistically ~100W for the 5v supply? The AOpen PSU delivers ~150W for the same supply.

So a new 750W is chasing an old 300W?

AOpen is rebranded Fortron (FSP) with strong 5V rail (30A) and combined 175W on 3,3 and 5 V. But I dont know, what you have inside: there was even Jamicon, Teapo or Capxon caps in FSP PSUs and these are not so reliable...

MattRocks wrote on 2026-04-16, 20:47:

What I have done is order my PSUs by cheapness, and then pair the more premium PSUs to CPUs matched by era: […]
Show full quote

What I have done is order my PSUs by cheapness, and then pair the more premium PSUs to CPUs matched by era:

AT/ATX transition MMX <-> 5V-focussed AOpen FSP300-60GT
AMD K6-2 ATX <-> 5v-focussed Delta DPS-400WB C
Athlon XP 2400+ <-> 5V/12V transition Antec TruePower Trio TP3-550
Athlon 64 2400+ <-> 12V-focussed Corsair RM750x

None of these systems, except possibly the Athlon XP 2400, need PSUs with really heavy 5V rails.

Really, the only systems that need 5V-heavy PSUs are Athlon/XP systems that don't have a 4-pin 12V CPU connector on the motherboard (most socket A motherboards), a few (very few) Pentium 4 systems that also lack a 4-pin 12V CPU connector on their motherboards, and any dual Pentium II/3 or Pentium Pro systems. For just about everything else, any PSU that can do 15 Amps on the 5V rail should be OK. Most single socket Pentium 2 and 3 systems will have a TDP of less than 60 Watts (typically less than 40W for the P3's), so 5V rail becomes relatively irrelevant.

With that said, the AOpen might be best relegated to the Athlon XP system if it doesn't have a 4-pin 12V CPU connector. Everything else should work in whatever other combination you like.
... well OK, some PSUs don't like to work with too little load and may not start, with the MMX and K6-2 systems most likely to be problematic here, as they tend to have pretty low power requirements.

MattRocks wrote on 2026-04-16, 20:47:

Having read the thread linked above, I feel somewhat... unsure... because I'm not an electrician.

Attached is a photo so you can see it has a fair number of expansion cards that I don't want to smoke. If the AOpen FPS PSU fails, is there any risk of damage to the components in my AT/ATX transition tower or is protecting?

TLDR about the AOpen PSU: it's 5VSB (5V standby) circuit could be potentially dangerous if any of its electrolytic caps fail. This can kill the motherboard and any attached hardware that directly draws power from the 5VSB rail... so possibly keyboards, mice, and/or LAN cards with WOL enabled. The cards that should be safe are anything not connected to the 5VSB rail - e.g. CPU, RAM, video cards, sound cards, HDDs, optical and floppy drives, just to name the most common stuff.
If recapped, though, that AOpen PSU should be a pretty solid 5V-heavy PSU.

cyclone3d wrote on 2026-04-16, 21:20:

If you get a high enough wattage ATX PSU, the 5v rails will generally be strong enough.

Yes, but then there's this possibility that some modern PSUs don't like to start with too low of a load, especially on the 12V rail.

Also, it really depends how the 3.3V and 5V rails are generated on modern PSUs.
Budget PSUs are more likely to use a group-regulated design, so they could still do very poorly when they are loaded more on the 5V rail than the 12V rail.
The better-designed PSUs will use buck-regulators to generate the 5V and 3.3V rails directly from the 12V rail... in which case, the PSU isn't going to care which rail you load it on the most, so long as you don't try to ask more than what the PSU is stated to provide on the label (in terms of the 3.3V and 5V rails.) Worth noting is that for such PSUs, there is usually no "combined" 3.3V and 5V power ratings to speak of, even if the label says there is. On PSUs that use buck-regulator cards/modules for the 3.3V and 5V rail, each of these can provide the current on the label independently from the other (since they both use the 12V rail for power.)

From MattRocks PSU list, I can tell you that the Antec Truepower and Delta are both ground-regulated units... so may not do too well on a 5V-heavy PC. However, they are also pretty sound PSUs and likely won't have problems starting even at very low loads.
The Corsair I'm not sure about. I have a similar one, though I don't remember if it was exactly an RM750.

MattRocks wrote on 2026-04-13, 17:03:

The one that blew up was a Mercury, which is one of the worst PSUs. I suspect Mercury may have been a high street store's own label for whatever was cheapest in the grey market.

If memory serves me right, Mercury was made by Leadman - the same company that makes the infamous PowMax PSUs. Real shoddy stuff at best. Even their old XT units were cutting corners like crazy... except the steel case for whatever reason.

momaka wrote on 2026-04-17, 20:21:

None of these systems, except possibly the Athlon XP 2400, need PSUs with really heavy 5V rails.

I don't think that is right. My understanding is that both Socket 7 systems are almost entirely powered by the 5v rail: CPU, motherboard, all the PCI cards (My AT case pictured has Matrox MMS, 3Dfx Voodoo, MPEG decoder, SBLink audio). SDRAM will come off the same rail too. When swapping HDD to SSD, there's almost no 12V demand in those platforms. I think that's min 100W steady load, and there needs to be some headroom.

The K6-2 built, while I'm less certain about final configuration, is only compatible with 3.3v AGP cards that draw from the 5v rail.

Both will lean heavily on the 5v supply and make very little use of the 12v supply.

Also, remember back in the day, the PSU requirements were given for how bad most of the trash PSU were.

Very old PSUs were rated for max temporary wattage, while any decent quality newer PSUs are rated for max continuous wattage.

For ATX PSUs I buy, I generally look for at least 80+ Bronze from good companies such as Seasonic. Other brands generally use various OEMs to manufacturers their PSUs so you need to find out who the OEM that made the specific model, and sometimes revision of that model is.

MattRocks wrote on 2026-04-17, 20:40:

I don't think that is right. My understanding is that both Socket 7 systems are almost entirely powered by the 5v rail: CPU, motherboard, and the PCI cards. When swapping HDD to SSD, there's almost no 12V demand in those platforms.

Yes, but their whole power draw is very small, so there's just no need for *any* rail to be very strong.
Just look various CPU TDPs for these and you will see.

A 166 MHz MMX, for example, is rated for about 16 Watts tops. Add another Watt or two for worst case VRM inefficiencty, and you're still at under 20W - that's less than 4 Amps on the 5V rail. A K6-2 500 is rated a mere 5 Watts more than the 166 MMX. Now add another 5W for RAM, and 10W tops for all other cards. You should be looking at 35-45 Watts max... which is still under 10 Amps from the 5V rail. Therefore, just about any PSU should technically be able to power these. The problem comes with certain group-regulated ATX v2.x PSUs, which just might not regulate properly due to the unbalanced 5V/12V rail loads. For such units, it won't matter how high you go with the power rating - the 5V/12V rail imbalance will always throw them off. FSP ATX v2.x compliant units tend to fall in that category. Delta PSUs tend to handle cross-loading much better, so that DPS-400WB of yours may not have any problems. Ditto for the Antec Truepower.

And this is why most socket 7 stuff can be powered with a PSU under 100 Watts easily.
If you don't believe me, buy a wall plug-in power meter and see for yourself. The draw from the wall (AC) will be slightly higher due to inefficiency from the PSU. In any case, you may just be surprised at how economical those old PCs are. Then again, if you look at the heatsink size in these old systems (or rather, how few chips have heatsinks), it shouldn't come as a surprise at all that their power draw is so low.