Yep, bad design or faulty.

I personally don't trust any power supplies except ones from very reputable companies like Seasonic.

For AT systems, I use an ATX to AT adapter.

Could you provide more details of the PSU? Type name and some amps it should be able to provide per power rail?
Some rebrands use the builders type name (like a HP or Dell branded pus with FSP type name)

Maybe you could open one up and take a picture from the inside? That way we can help telling you whether it's a quantity PSU.
Of course you must be sure you'll want that, as opening it up will probably void any kind of warranty (if you have that).

Or if you can look inside thru the holes, you might see if all party's are mounted on the PCB (and none let away) and if the great sinks seem to be decent sized. That's a pretty ok way to see if it's probably ok quality.

It's a known easy way to cheap out a PSU by shrinking cooling sinks and to let away party's that are then bypassed by a little wire.

If all party's are there and seems decent, that's where my knowledge ends.

I don't think those Athena AT units are very good, at least from what I've heard, so pictures of the inside would definitely help. The Startech ones, on the other hand are quite reliable... the protection on those saved me from an incorrectly plugged P8/P9 connector the other day.

Another reason may be that the PSU requires a minimum +12V load to regulate properly, can you try adding a mechanical HDD to the mix and checking if that improves the voltages?

waterbeesje wrote on 2021-09-21, 18:37:

HP or Dell branded pus

I loved that 👍

TheMobRules wrote on 2021-09-21, 19:04:

Another reason may be that the PSU requires a minimum +12V load to regulate properly, can you try adding a mechanical HDD to the mix and checking if that improves the voltages?

Yes, it is that dreaded crossloading problem.
When one rail (here +5V) is loaded too heavily in relation to the +12V, the latter goes up.

Either replace with a good AT PSU (preferably Seasonic), a modern PSU that creates +5V etc from the main rail (e.g. 12V) using DC-DC converters or add dummy loads (HDD, resistors, loads of LEDs with resistors or whatever) until voltages stay in range.
If you add a HDD, make sure you take one which isn't too energy saving, e.g. a 3.5" or even 5.25", which draws at least 0.6-0.8A, the more, the better 😀

cyclone3d wrote on 2021-09-21, 17:44:

Yep, bad design or faulty.

I personally don't trust any power supplies except ones from very reputable companies like Seasonic.

For AT systems, I use an ATX to AT adapter.

Hi, thanks for the response! That would be a great solution and I would feel more comfortable running a modern, quality power supply - especially from a brand like Seasonic.

I initially steered away from this as I was worried about the lack of -5V on modern power supplies. But, maybe I'm unnecessarily worried about that. I think I now understand that only the ISA bus potentially used -5V, and the only ISA card I'm looking at is either an SB16 or AWE64 which I believe don't use the -5V power. Does that sound right?

Am I therefore ok to just run without -5V and just need to ensure that the ATX power supply has a strong enough 5V rail? Any other specs I should look for in the power supply?

Thanks again!

TheMobRules wrote on 2021-09-21, 19:04:

I don't think those Athena AT units are very good, at least from what I've heard, so pictures of the inside would definitely help. The Startech ones, on the other hand are quite reliable... the protection on those saved me from an incorrectly plugged P8/P9 connector the other day.

Another reason may be that the PSU requires a minimum +12V load to regulate properly, can you try adding a mechanical HDD to the mix and checking if that improves the voltages?

Hi, thank you and retardware for the response!

I’ve added a Seagate Barracuda which draws +5V@0.85A and +12V@0.75A and the voltage on the 12V rail reduced from 12.99V to 12.91V. Not a huge change, but it sounds like confirms the cross-loading design issue?

Photos and details on the power supply forthcoming once I get it out of the system.

waterbeesje wrote on 2021-09-21, 18:37:

Could you provide more details of the PSU? Type name and some amps it should be able to provide per power rail? Some rebrands us […]
Show full quote

Could you provide more details of the PSU? Type name and some amps it should be able to provide per power rail?
Some rebrands use the builders type name (like a HP or Dell branded pus with FSP type name)

Maybe you could open one up and take a picture from the inside? That way we can help telling you whether it's a quantity PSU.
Of course you must be sure you'll want that, as opening it up will probably void any kind of warranty (if you have that).

Or if you can look inside thru the holes, you might see if all party's are mounted on the PCB (and none let away) and if the great sinks seem to be decent sized. That's a pretty ok way to see if it's probably ok quality.

It's a known easy way to cheap out a PSU by shrinking cooling sinks and to let away party's that are then bypassed by a little wire.

If all party's are there and seems decent, that's where my knowledge ends.

Hi, thank you for the response!

The power supply is an Athena Power AT40 400W power supply. The rails provide:

+12V@18A
+5V@32A
-5V@0.5A
-12V@0.5A

I'm attaching photos of the unit. I really have no idea what I'm looking at, but would be interested in comments on the quality of the build.

Thanks again!

Ugh that looks like FSP :vomit:
No filter and fan of the cheapest sort.
That is not what I like or even would consider using myself.

Asides of that, just the current draw of the inactive HDD of maybe 1/3 to 1/2A on 12V probably doesn't constitute a sufficient load for this PSU.
You need something heavier.
Or just a good PSU.

Yeah.. I wouldn't be using that PSU.

retardware wrote on 2021-09-22, 01:11:

Ugh that looks like FSP :vomit: No filter and fan of the cheapest sort. That is not what I like or even would consider using my […]
Show full quote

Ugh that looks like FSP :vomit:
No filter and fan of the cheapest sort.
That is not what I like or even would consider using myself.

Asides of that, just the current draw of the inactive HDD of maybe 1/3 to 1/2A on 12V probably doesn't constitute a sufficient load for this PSU.
You need something heavier.
Or just a good PSU.

That good huh? Sigh, I really wanted to like these as I was so happy to see someone manufacturing new AT power supplies.

Just as a bit of a tangent, I thought FSP was generally well regarded. Did I get that wrong, or does it vary between models?

I'll see what I can find in terms of a good Seasonic power supply with a strong 5V rail and an ATX to AT adapter.

Thanks again!

cyclone3d wrote on 2021-09-22, 01:14:

Yeah.. I wouldn't be using that PSU.

Thanks! I'm disappointed about the PSU, but happy to be steered away from it now rather than end up in tears down the road.

joed96 wrote on 2021-09-22, 01:26:

Just as a bit of a tangent, I thought FSP was generally well regarded. Did I get that wrong, or does it vary between models?

My personal impression is that FSP is best regarded by people who did not do test bench examinations.
I originally considered FSP an acceptable brand until I did my tests. What I found made me dislike FSP very much. Bad design, dangerous construction (on some of their PSUs one can touch live with the fingernail) and other questionable things,

retardware wrote on 2021-09-22, 03:07:

joed96 wrote on 2021-09-22, 01:26:

Just as a bit of a tangent, I thought FSP was generally well regarded. Did I get that wrong, or does it vary between models?

My personal impression is that FSP is best regarded by people who did not do test bench examinations.
I originally considered FSP an acceptable brand until I did my tests. What I found made me dislike FSP very much. Bad design, dangerous construction (on some of their PSUs one can touch live with the fingernail) and other questionable things,

Thanks for sharing, very insightful! It certainly helps to be more aware.

joed96 wrote on 2021-09-22, 00:58:

Hi, thank you for the response! […]
Show full quote

Hi, thank you for the response!

The power supply is an Athena Power AT40 400W power supply. The rails provide:

+12V@18A
+5V@32A
-5V@0.5A
-12V@0.5A

I'm attaching photos of the unit. I really have no idea what I'm looking at, but would be interested in comments on the quality of the build.

Thanks again!

The secondary side of that PSU is brutally undersized (side opposite of the large primary caps) Should be a LOT more capacitance on that secondary side, at least for this particular design at this particular wattage.

I'm not going to turn around and say the design if particularly bad ... I see this design a lot and a lot of computers operate fine on PSUs of this design. But this particular example is no good.

And if the group regulation hasn't been tuned properly for 5V heavy applications, you get what you are seeing with the 12V shooting up as the PSU inadvertently adds more energy to the 12V circuit at the same time to delivering more energy to keep the 5V rail propped up.

First of all, that PSU is not FSP at all. While FSP is not a top tier manufacturer, their designs are at the very least decent and their units are packed with components and quite heavy, that one is full of air.

Also, that power supply is just a regular bad quality ATX repurposed as AT, you can even see the empty spots for the 3.3V output components and rectifier.

Finally, old 5V based hardware is much more tolerant of ripple than newer low voltage components, that's why so much stuff from the 90's still survives to this day despite being run with power supplies that would be considered inadequate today. In fact, even the earlier ATX specs allowed up to 120mV of ripple on the +12V line, which was later reduced to 60mV. So just get something from a known manufacturer that is in spec and call it a day, no need to overthink it.

canthearu wrote on 2021-09-22, 05:10:

And if the group regulation hasn't been tuned properly for 5V heavy applications, you get what you are seeing with the 12V shooting up as the PSU inadvertently adds more energy to the 12V circuit at the same time to delivering more energy to keep the 5V rail propped up.

TheMobRules wrote on 2021-09-22, 05:15:

So just get something from a known manufacturer that is in spec and call it a day, no need to overthink it.

Thanks you both canthearu and TheMobRules. That definitely helps me understand why my voltage was going up and that I clearly want to stay away from that power supply. And you were right, I was massively overthinking the -5V rail that I don't even need.

I went down to the market and got a Seasonic Focus GX-1000 - overkill over all, but it gives me 25A on the 5V rail and fits in my case. I'm getting 5.066V and 12.10V with this power supply, and I'm much more confident in my setup.

Thanks again everyone!

For what it's worth, this was my experience with Athena AT-40 PSU: And then there was white smoke...

TheMobRules wrote on 2021-09-22, 05:15:

First of all, that PSU is not FSP at all. While FSP is not a top tier manufacturer, their designs are at the very least decent and their units are packed with components and quite heavy, that one is full of air.

Also, that power supply is just a regular bad quality ATX repurposed as AT, you can even see the empty spots for the 3.3V output components and rectifier.

To me it looks like a custom-cut-down FSP PSU.
The middle cooler that touches the case is characteristic for the FSP design. On some models, it is attached to the case with a screw, on others it is insulated from the case using a thin piece of soft thermo foil that one can pierce with the fingernail.
Another characteristics is the mains electrolytics which are surrounded by hot stuff in a sparsely-vented part of the board.
However, I didn't spot some other typical FSP "goodies" like electrolytics wrapped in shrink tubing to "insulate"/"protect" them from nearby hot transformers and coolers.
These have probably cut down like the filters for cost-saving.

TheMobRules wrote on 2021-09-22, 05:15:

Finally, old 5V based hardware is much more tolerant of ripple than newer low voltage components, that's why so much stuff from the 90's still survives to this day despite being run with power supplies that would be considered inadequate today.

Not so sure about this. The early DC-DC regulators had to work much harder when +5V was down substantially.
I have a 1999 AT form factor Tualatin board with bulged electrolytics and replaced CPU mosfets from Russia, and I would be so curious about the PSU it was being operated on before that PC was being dismantled for parts.

TheMobRules wrote on 2021-09-22, 05:15:

In fact, even the earlier ATX specs allowed up to 120mV of ripple on the +12V line, which was later reduced to 60mV. So just get something from a known manufacturer that is in spec and call it a day, no need to overthink it.

Yes I agree with you, especially as the modern trend is to derive +5 and +3.3V from 12V via separate/additional DC-DC converters. This makes the annoying crossloading problem nonexistent.