I have never had problems with various modern 20A 5v units with my Athlon / Athlon XP setups over the years.

A friend of mine had a more modern PSU die on him when he used it to power on one of his Athlon XP systems and it destroyed it's 5v rail, even though it had listed something like 25A on the 5v line. It was not some El Cheapo PSU, it should've been decent, was around 600W and had powered his A64x2 6000+ without any issues. I can't get him on the phone so I don't know exactly which PSU failed on him, but RMA-ing it took months even though it was still in warranty. The 12v rail still worked though 🤣

Some s462 boards are using the 12V P4 connector. These should have no issues even with modern PSU.

havli wrote:

Some s462 boards are using the 12V P4 connector. These should have no issues even with modern PSU.

This is correct. All Athlon XP boards this friend of mine used, didn't have this connector though (all of them were VIA chipsetted boards, KT400 iirc). Iirc NForce boards had the P4 power connector more often, though the KT600 board I used to use a lot did actually have the P4 power connector, but it was a Gigabyte OEM board made for Fujitsu Siemens (maybe those are different).

Yeah we've discussed this Athlon PSU stuff absolutely to death. If I have a PSU blow up in one of these decades, I'll let you know. I've been using FSP and Seasonic built PSUs.

I really don't like the idea of telling people to buy circa 2001 PSUs that probably need to be refurbed. If the person is knowledgeable about PSUs construction though, then have at it.

Tetrium wrote:

A friend of mine had a more modern PSU die on him when he used it to power on one of his Athlon XP systems and it destroyed it's 5v rail, even though it had listed something like 25A on the 5v line. It was not some El Cheapo PSU, it should've been decent, was around 600W and had powered his A64x2 6000+ without any issues ...l

There are two potential problems when using modern '+12V heavy' PSUs to power old '+5V heavy' comps, like Athlon XP without 'P4' connector.

· +5V rail rating. If PSU has say one 15A rectifier on +5V (15x5= 75W) and comp demands 100W from +5V rail at full steam, BANG.

. Voltage regulation. Most budget PSUs are group-regulated (as opposed to independent regulation). Long story short, this means they can't handle well a combination of heavy load on the +5V rail and light load on the +12V rail:

http://www.techpowerup.com/reviews/Corsair/CX430_V2/5.html

^ In the crossload test CL1, a combination of light draw from +12V (only 2A or 24W), but heavy draw from +5V (13A or 75W) and from 3.3V (15A or 50W), the group-regulated Corsair CX430 V2 can't cope and +5V rail voltage drops below ATX specification (red rectangle). Ouch.

Now that's an extreme torture test from a PSU reviewer, not likely to be reproduced on a real world comp, but the potential for problems is there, check +5V voltage with a multimeter to be safe.

There are things that draw extra from +12V and will help on voltage regulation when modern '+12V heavy' group regulated PSUs are powering old '+5V heavy' comps:

- Fans, HDDs, optical drives, some GPUs.

- Dummy loads, like 12V 21W car bulbs, or heavy duty resistors (this trick is as old as the PC itself 🤣 ).

swaaye wrote:

Yeah we've discussed this Athlon PSU stuff absolutely to death. If I have a PSU blow up in one of these decades, I'll let you know. I've been using FSP and Seasonic built PSUs.

I really don't like the idea of telling people to buy circa 2001 PSUs that probably need to be refurbed. If the person is knowledgeable about PSUs construction though, then have at it.

I agree with you completely. Personally I think PCBONEZ's and TELVM's knowledge about PSU's is far superior to my own and to tell you the truth, I was actually using those old PSU's you mentioned (the vast majority being FSP's actually 🤣) but since then these old PSU's have aged even more over the years and perhaps seen a lot of extra use as a bonus.

Skyscraper wrote:

sgt76 wrote:

Athlon 64, I've had the worst luck with socket 462 systems... dunno why, curse of the hardware gods mayve. 😢

I agree its better to go with Socket-754 or Socket-939 for a Windows XP build but if going with Socket-A my tip is to avoid the Asus A7N8X as it uses a 5V power design for the CPU.

My Asus A7N8X Deluxe killed my rare 2333(166) MHz version of the Barton XP 3200+ when restarting during a benchmark session. The temperature was OK and the voltage set was only ~1.8V so there must have been a voltage spike or something like that which killed the CPU.

That's really very sad 🙁

havli wrote:

Some s462 boards are using the 12V P4 connector. These should have no issues even with modern PSU.

Those do as well. Even if the 12v connector is present on the board, most (especially popular NF2 models) still use the 5V rail to power the CPU. The only socket A boards that draw CPU power from the 12V rail I know of are the ABIT NF7, the Gigabyte 7NNXP and the MSI KT880 Delta series.

There is a similar issue on the Intel side.
Intel recommended that CPU power be moved to +12v for socket 370 --
-- but (with the possible exception of SOME dual CPU server boards), no one I know of (not even Intel) actually did it until their P4 boards.
.
There may even be some goof-ball early P4 boards that power the CPU with +5v but i don't recall ever seeing one.
.

The actual purpose of the famous "P4 connector" is to reduce the current loading on the traces in the motherboard - by adding more traces for +12v to route through.
Secondarily, it adds more pins for +12v to pass through. Pins also have current limits.
.
Some mobo designs took things too far and split the +12v from the P4 connector from the +12v in the 20/24 pin into two independent +12v rails on the mobo itself.
That was a design error.
It basically defeats the purpose of adding the P4 connector by reducing the trace area and number of pins the CPU power passes through.
.
.
Just FYI:
The very early standard pins were good for something like 6 or 7 amps/pin. (I don't remember exactly. Those faded away long ago.)
Currently the standard pins are good for 9 amps/pin and the heavy duty are good for 13 amps/pin.
There is a wire size requirement to the max current rating from the HD pins. 16 ga wire must be used.
- Those numbers are specifically for Molex brand pins. It varies a little bit with other manufactures but it's all pretty close.
.

PCBONEZ wrote:

There is a similar issue on the Intel side. Intel recommended that CPU power be moved to +12v for socket 370 -- -- but (with the […]
Show full quote

There is a similar issue on the Intel side.
Intel recommended that CPU power be moved to +12v for socket 370 --
-- but (with the possible exception of SOME dual CPU server boards), no one I know of (not even Intel) actually did it until their P4 boards.
.
There may even be some goof-ball early P4 boards that power the CPU with +5v but i don't recall ever seeing one.
.

I think there were! I recall reading an early P4 review where they tried powering the system (an early S423 board) without the P4 connector plugged in, and it worked. Though IIRC this board also had one of those 6-pin aux connectors as well, so that may have been helping out. I can't remember which site that was on.

Standard Def Steve wrote:

PCBONEZ wrote:

There is a similar issue on the Intel side. Intel recommended that CPU power be moved to +12v for socket 370 -- -- but (with the […]
Show full quote

There is a similar issue on the Intel side.
Intel recommended that CPU power be moved to +12v for socket 370 --
-- but (with the possible exception of SOME dual CPU server boards), no one I know of (not even Intel) actually did it until their P4 boards.
.
There may even be some goof-ball early P4 boards that power the CPU with +5v but i don't recall ever seeing one.
.

I think there were! I recall reading an early P4 review where they tried powering the system (an early S423 board) without the P4 connector plugged in, and it worked. Though IIRC this board also had one of those 6-pin aux connectors as well, so that may have been helping out. I can't remember which site that was on.

Even without an aux connector there is one +12v line in the 20-pin ATX connector. (An early board probably had a 20-pin and probably without the HD pins.)
That's enough to give the CPU VRM it's +12v but with that much power going through one trace (and 1 pin) I wouldn't expect it to last very before something fried.
.
If that aux connector was similar to more modern server boards (2x3 or 2x4 pins with +12v and ground pins) it was usually an either-or situation between the "aux" and the "P4" connectors.
(For older boards anyway.)
If you start getting into later Xeon dual CPU boards, -some- of those have 3 aux connectors. One for each CPU and another for RAM. (RAM has it's own VRM on those boards.)
.