TELVM wrote on 2026-07-01, 16:24:

^ Group-regulated modern(ish) PSUs, designed for "+12V heavy / minor rails weak" systems in the post-P4 era. […]
Show full quote

^ Group-regulated modern(ish) PSUs, designed for "+12V heavy / minor rails weak" systems in the post-P4 era.

Under opposite "+5V heavy / +12V weak" demands, the group-regulated topology can't keep voltages within ATX spec (+5V plummets down, +12V skyrockets upwards).

In this review of a ~2019 Corsair VS650, also group-regulated ...

... look at the Crossload Test 1 ("+12V weak, +5V heavy" draw):

^ Both +12V & +5V voltages are FUBAR.

.

Group-regulated Corsair VS450 defeated by Socket-A system in a Phil's Computer Lab video:

Can your PSU handle Athlon or Duron Socket A processors?

^ If he had used a multimeter we'd see +5V voltage dropping down as he inserts components of progressively higher +5V draw.

TELVM wrote on 2026-07-01, 16:24:

^ That's why modern PSUs with DC-DC regulated minor rails are better suited for retro-PCs than group-regulated ones, voltages are rock-steady regardless of cross-loading.

I don't think that's the difference. The chinesium unit worked fine in a 2008 PC with CPU connector, and using that corsair in a socket A system without CPU connector the 5v rail would still drop to 4.60v.
Pretty sure the main difference here is the lack of 12v CPU connector on that mobo. Only old PSUs would work well on that. But if a 462 mobo does have the P4 CPU connector, then any modern PSU will do.

TELVM wrote on 2026-07-01, 16:24:

^ That's not good at all. The +5V DC-DC FETs lack heatsinks, and with the fan stopped, and no air cooling, they may overheat and die, see here.

Yeah, because the load on the 12V isn't enough to trigger the fan. The CPU is around 60W, the rest of the 12v line would pull a few watts, so we are well below 100W. I don't know how much wattage is being pulled from the 5V rail, considering the GPU is just an FX 5600.
Is there a way to bypass the smart fan controller and just keep that fan always spinning?

EDIT: There isn't a way, but corsair own manuals says that, for the RM650x (the PSU I'm using in that socket A system), the fans spins up from 260W loads and upwards.
No way in hell that system is ever gonna pull 260W. I better put a fan blowing air right into it then.

Mondodimotori wrote on 2026-07-02, 07:55:

... Pretty sure the main difference here is the lack of 12v CPU connector on that mobo. Only old PSUs would work well on that. But if a 462 mobo does have the P4 CPU connector, then any modern PSU will do ...

^ Yep, that's the problem with modern(ish) group-regulated PSUs, their design expects the larger load on +12V, and only relatively small loads on +5V & +3.3V (= a crossloading heavily tilted on the +12V side of the scale).

A 462 mobo with "P4" connector, where the CPU draws +12V, satisfies the expectations of a PSU of this kind.

In the table a couple posts above, at "Crossload 2", notice how the VS650 has no problem handling 52A (624W) on +12V, as long as the minor rails draw little juice (just 1A each).

.

Any decent group-regulated PSU from just before "the P4 era" was designed the opposite way, expecting heavy draw from +5V and relatively little from +12V (= a crossloading heavily tilted on the +5V side of the scale). An extreme example:

Conversely, if we'd try to use such a "+5V strong / +12V weak" classic PSU to power a modern "+12V hungry" system, it would be defeated. Either it wouldn't start, or the +12V rectifiers would go KABOOM at once.

Mondodimotori wrote on 2026-07-02, 07:55:

... the fans spins up from 260W loads and upwards.
No way in hell that system is ever gonna pull 260W. I better put a fan blowing air right into it then. ...

The scintillating "ZERO RPM FAN MODE":

A large 135mm fan, that can move tons of cooling air, stopped at 0 rpm. Ain't that smart?

Other PSUs have a switch to cancel the nonsense, but it looks like the RM650x lacks one.

shevalier wrote on 2026-06-29, 12:17:

Even if we disregard the fact that the primary-side capacitors are rated for 250 watts. […]
Show full quote

TELVM wrote on 2026-06-29, 10:16:

PSU Boogaloo - Testing a Lead Power LP-650 ATX Power Supply
.

Even if we disregard the fact that the primary-side capacitors are rated for 250 watts.

Since 2001, the sale of power supplies with a power rating exceeding 75 watts that do not feature a any type of power factor corrector (PFC) has been banned in the EU.
Since 2014, the sale of power supply units rated at over 75 watts without active PFC has been prohibited in the EU.

Where did the author of this video find this marvellous work of art?

That's the kind of stuff you get from Aliexpress, eBay, or Amazon.

I like this one. Their "heatsinks" are aluminum-backed PCBs with the parts soldered to them. I don't know who thought that was a good idea, but someone must have had a ton of surplus PCB material from a canceled LED lighting project or something.
https://www.youtube.com/watch?v=3TyQEaNFjuc

On the other hand, Dell managed to get Chicony to build them power supplies without PFC in 2018. My parents have one. It's 240W with a voltage selector switch (and I don't know if it's the model with the underrated snubber cap).

TELVM wrote on 2026-07-02, 18:54:

The scintillating "ZERO RPM FAN MODE": […]
Show full quote

The scintillating "ZERO RPM FAN MODE":

A large 135mm fan, that can move tons of cooling air, stopped at 0 rpm. Ain't that smart?

Other PSUs have a switch to cancel the nonsense, but it looks like the RM650x lacks one.

Yup.
In any modernish PC, even the first PCIe from 20 years ago, reaching a power pull of 240 watts under load wasn't hard. But before that? Yeah, it competely fucks up with case airflow too.
I'll just add another fan that will push air through the PSUs and out of it, just in case. But these "RMx" units from 2016 and after (white lable) should be sturdier and not buckle under intense 5v load.
I even doubt I'll load the 5v rail that much with an XP 2600+ and a P4 connector.
But for the system without that connector? a chuncky Enermax with the big fan always spinning at max speed, keeping both the PSU and the rest of the PC nice and cool.

lti wrote on 2026-07-02, 19:46:

... I like this one. Their "heatsinks" are aluminum-backed PCBs with the parts soldered to them. I don't know who thought that was a good idea, but someone must have had a ton of surplus PCB material from a canceled LED lighting project or something.

https://www.youtube.com/watch?v=3TyQEaNFjuc ...

Wow! Those priceless heatsinks deserve a place of privilege in the Gutless Wonder Hall of Fame!

They should brag about them on the box:

"Featuring stealth heatsinks, with extremely low radar cross section, near invisible optical signature, and presenting a minimal target to enfilading fire - Endless victory for the uncompromising gamer" 😀 .

Mondodimotori wrote on 2026-07-02, 20:15:

... I'll just add another fan that will push air through the PSUs and out of it, just in case ...

You may disconnect the OEM 135mm fan from its header, then connect it directly to +12V (always full rpm, noisy), or +5V (always ~500 or so rpm, quiet).

TELVM wrote on 2026-07-03, 00:12:

You may disconnect the OEM 135mm fan from its header, then connect it directly to +12V (always full rpm, noisy), or +5V (always ~500 or so rpm, quiet).

I'm not really a fan (eheh, get it, fan) of messing inside a PSU.
But what's bugging me is not that, it's another board I have that, despite reporting in spec voltages in BIOS, once in windows, no matter if Everest, AIDA64 or Hardware Info, I get 5.5v on the 5v rail and well over 15V on the 12v rail. It's not ever a matter with a bad PSU, since it happens with two known working PSUs and only on this board, a K8V-MX from ASUS.
Is it just bad readings from the sensors once in windows? Because, correct me if I'm wrong, but such voltages on the 12v rail should fry a PC, right? And in BIOS they all are perfectly within spec.

Mondodimotori wrote on 2026-07-05, 20:10:

... Is it just bad readings from the sensors once in windows? ...

Yep, that's it. Software voltage readings can't be trusted.

Mondodimotori wrote on 2026-07-05, 20:10:

... in BIOS they all are perfectly within spec.

Voltages should be OK then.

To be completely sure, a multimeter is a cheap and handy tool.

How to Manually Test a Power Supply With a Multimeter

Are you going to be testing them under load with a multimeter though?

The other thing you can't really test for is ripple which can wreck havoc on not only stability but also on the life of the components being powered.

cyclone3d wrote on 2026-07-06, 17:50:

Are you going to be testing them under load with a multimeter though? ...

Sure, why not?

cyclone3d wrote on 2026-07-06, 17:50:

... The other thing you can't really test for is ripple which can wreck havoc on not only stability but also on the life of the components being powered.

^ Very true. You need an oscilloscope for that, not a cheap investment for a hobbyist.

.

FUBAR output caps causing murderous 500mV of ripple (TEN times the allowance in ATX specs) on the +3.3V rail of a PSU:

Replacing Failed PSU Electrolytic Caps. Can we fix the ripple?

That makes me want to test all my PSUs

ATX specs for voltage and ripple:

just a reminder that PSUs can mess with drives (DVDs, HDDs, SSDs, whatever!) and other components, figured out that today so keep in mind just in case if you think your motherboard can be faulty or reset everytime.

TELVM wrote on 2026-07-07, 10:40:

ATX specs for voltage and ripple: […]
Show full quote

ATX specs for voltage and ripple:

There's just one problem.
It's important to clearly distinguish ripple from transient.
Ripple noise "voltage" is the "change in current"*"ESR of the capacitors".
And yes, it's limited to millivolts.
Transient is the change in voltage due to a change in load, i.e., how fast the power supply responds.
And there, it's still the same +/- 5%, i.e., hundreds of mV (for example, 0,6V for +12V rail).
And on an oscilloscope, you see their sum, and it's hard to tell one from the other. 🙁

For example, the 650W PSH from CWT is branded as Chieftec.
Under a 350W load, it has a ripple of 20mV and some nasty 0.1V interference with the main supply 100Hz frequency (which I didn't really look for), because it's completely within the specifications.

shevalier wrote on 2026-07-08, 06:34:

... Transient is the change in voltage due to a change in load, i.e., how fast the power supply responds ...

Example of non-stellar transient response on +3.3V line:

XPG Pylon 750 W Review

Interesting video about fixing a classic cheap (TL494 controller) PSU:

Let's fix this DEAD ATX power supply!

Healthy 35A rated +5V line:

However most of the caps are FUBAR, to the point that the controller can't even get enough voltage to start up.

^ ".... not unexpected with such prestigious brands been used ... " 😀 😀 😀 😀 😀

.

After recapping he concludes: "It seems to be working fine, 12 volts is fine, 5 volts is fine ..."

But from 23:53 onwards notice how, as he increases the load amp after amp on the +5V line, +5V voltage progressively drops, while +12V climbs up simultaneously, until at 7A (35W) load on +5V they are at 4.64V (out of ATX spec) and 12.49V respectively:

^ That's the group-regulated topology being defeated by a "+5V heavy / +12V light" crossloading.

Even on a supposedly "~period correct, +5V heavy", though of not so great quality, PSU.

Pity he didn't put load on 12V rail to simulate real load and figure out how much is needed to get it back in ATX spec. With 2 magnetic HDDs and a few case fans you can probably get to about 1.5A. To consume more you would probably use 12V case led strips. Total load on 12V about 2A.

I was reading reviews of the Seasonic S12III, and some of them affirm that it is group-regulated.

But looking at pics of the guts I wasn't so sure ...

Then I found this very thorough review in vietnamese.

Google translated excerpts:

" ... Surely this is the point that users expect to upgrade compared to the S12II version.

With PSUs designed using a DC to DC circuit to create two lines 5V and 3.3V from the 12V line, the purpose is to use the main transformer to output only one 12V line to solve the problem of ensuring voltage stability when loading asymmetrically between the power lines that "traditional" (group regulated) designed PSUs have because the main transformer outputs two 12V and 5V lines at the same time ..."

"... Seasonic uses another technology that is simpler but also provides the same voltage stability as VRMs DC to DC, the "dual magnetic amplifier" Aka Dual Mag, which they used before on the S12/M12 series (model: SS-xx0HT), but at that time (about 13~14 years ago) generations of computers (PCs) still used 5V and 3.3V lines quite a lot and Dual Mag did not really bring a difference. especially while production costs will be higher with the design used on the S12II.

The way the Dual Mag works is similar to the way a "traditional" PSU creates a 3.3V line from a 12V line. With Dual Mag, the 5V line is also created similarly. Through a small low-voltage inductor connected directly to the output of the main transformer ..."

Unfortunately the only crossload test he did was "+12V heavy / +5V & +3.3V light", not the reciprocal that would be of interest to us.

But I believe he's right and the S12III is indy-regulated, the old style. Which could make it a decent option for powering (not too hungry) retrocomps.

What do you guys think?

TLDR can you people tell me PSUs with good 3.3V-5V lines again? My problem is that on my slot 1 setup either my IDE-SATA adapter messed up or my 5V line drops to 4.916 and cannot properly mess with adapter/SSD (rarely detects it)

Takino-42 wrote on 2026-07-19, 11:31:

TLDR can you people tell me PSUs with good 3.3V-5V lines again? My problem is that on my slot 1 setup either my IDE-SATA adapter messed up or my 5V line drops to 4.916 and cannot properly mess with adapter/SSD (rarely detects it)

If 4.9V are an issue for that adapter, it's garbage imho. I wouldn't blame the PSU for that.

Takino-42 wrote on 2026-07-19, 11:31:

TLDR can you people tell me PSUs with good 3.3V-5V lines again? ... slot 1 setup ...

A typical single-CPU Slot 1 rig draws less than 100W overall. So any modern PSU rated for ~20A / ~20A on +5V / +3.3V will do fine, provided it has DC-DC rectification for the minor rails (to avoid "+5V heavy / +12V light" crossload problems). And provided the rig doesn't need a -5V rail (modern PSUs lack -5V).

Just two random examples of budget, but decent, modern PSUs that will work OK on a typical single-CPU Slot 1 system:

XPG Pylon 450W (year 2022 - ATX12V v2.53, EPS 2.92)

Corsair CX450M (year 2022 - ATX12V v2.4, EPS 2.92)

^ Unlike many budget PSUs that come with bottom-of-the-barrel crapacitors to reduce costs, these two sport a jap bulk cap, and mid-tier Elites for output filtering.

Takino-42 wrote on 2026-07-19, 11:31:

... my IDE-SATA adapter messed up or my 5V line drops to 4.916 and cannot properly mess with adapter/SSD (rarely detects it)

^ 4.916V is well inside ATX specs, I don't think +5V voltage is the problem here.