New PSU for a dual 370

[zandrax]

Hi everybody.

I've assembled a dual Pentium 3 server and I'm looking for a decent 350-400 W power supply.

At the moment, the server is based on:
- hp tc3100 motherboard (a customized Asus TR-DLS, with a proprietary diagnostic connector, a different bios, onboard Lsi SCSI adapter and only one Ide channel);
- dual P3 Coppermine 1000EB;
- one 512 MB PC133 ECC Registered module;
- a BenQ combo drive;
- a Samsung PL40 40 GB Ide hard disk.

The Samsung unit is temporary: I'm planning to join 3 o 4 Scsi hard disk in a raid 5 array, in order to turn the server into a home nas.

Intially I used a 400 W rated Allied psu: it run fine for a while until the computer suddently rebooted after a 4 hours Memtest session. Since I could firmly exclude blackouts, overheating issues [the heat sinks and the air were barely warm], mistakes with cabling and so on, I opened the psu: I think everybody here knows the [ironic] "high quality build" [/ironic] of Deer/Allied products, but I was simply astonished in discovering my old AT 150w Hipro had same capacity, better brand capacitors, thicker cables and a more ordinated layout. What a crap!

Eventually I replaced the Allied with the other Atx psu I have, an old 200 w Power Win I have used since 1998: it seems it's running fine (just passed a 5 hours Memtest by now), but in the middle run I'll need a more powerful one.

I live in Italy and, due to import and value added taxes, I'm unable to buy outside Europe: Sparkle, Zippy, Topfield and others are almost forbidden due to high prices.
The Corsair VX450 [70 € or $100] divides the "good" from the "bad" psus, but its 5 and 3.3 V rails are a bit weak; for less money, only Fortron emerges from the serious crap (Allied, Tagan, LC Power and others).
Among the rare Oems, I've found two Newton Power: the NPS-400AB A and B. Both are fairly priced, claim 26A on the 5v rail and 24A on the 3.3 one and are new: the A model has an EPS 8 pin connector, has 18A on the first and 10A on the second 12V rail; the B one, instead, has an EPS 4 pin, a 6 pin PCIe and two 15A 12v rails.

I would be interested in the B one: any suggestions?

Thanks in advance,
Zandrax

NPS-400AB A label:


NPS-400AB B:

[willawake]

Re: New PSU for a dual 370

what the hell is newton power????

[zandrax]

Re: New PSU for a dual 370

what the hell is newton power????

According to this thread, Newton Power Ltd seems a Delta subsidary and a Dell Oem.
In fact, every NPS-xxx I find seems to power a Dell pc.
Unfortunately I don't have any information about their quality and reliability, so I asked here.

Zandrax

[willawake]

Re: New PSU for a dual 370

hmm ok interesting, i missed that thread

i think you need the 4 pin eps?
pity they have dual rails, it would be better to have single fat rail for your setup

still should be ok

[zandrax]

Re: New PSU for a dual 370

i think you need the 4 pin eps?

I don't need it at all: the TR-DLS has only an Atx 24 pin and accepts even an old 20 pin, otherwise, my old psu couldn't sustain the Memtest (over 8 hours by now) and the others torture test I've planned.
I simply choose the B model because its 12v rails are more balanced, just that.

pity they have dual rails, it would be better to have single fat rail for your setup

still should be ok

I agree, but nobody could sell a new psu in Europe with more tha 20A on a single 12v rail nowadays, because it would violate the EN 60950.
According to the Newton UL page, the 400AB comes near to this limit, about 19A per rail.
However, I don't believe every psu has a single 12v transformer for every 12v line: more likely, the line is split after the transformer.

After the purchase, I'll post some photos of it.

Thanks,
Zandrax

[PCBONEZ]

Re: New PSU for a dual 370

Splice a jumper wire between a wire on 12v1 and a wire on 12v2.
TA DA! - One rail.
Use one jumper per each approx 8 amps you might be pulling across it.

Better yet solder a 14 gauge jumper on the PCB from 12v1 to 12v2 connection points.

~~~

And before all that check if the rails are internally connected already with an ohm meter through the wires. [Many that say they are two rails on the output really aren't.]

~~~

The Newton Power PSU's Ive seen were decent.

In fact I have one out of an old Gateway apart right now.
Checking if the caps in there might be the problem.
Answer: NOT.
Its full of Chemicon and Rubycon caps.
No crap caps at all.

The biggest problem with them is the OEM's tend to choose the smallest one that they can possibly get away with so you end up replacing them for even a little bit of an upgrade.

.

[zandrax]

Re: New PSU for a dual 370

Splice a jumper wire between a wire on 12v1 and a wire on 12v2.
TA DA! - One rail.
Use one jumper per each approx 8 amps you might be pulling across it.

Better yet solder a 14 gauge jumper on the PCB from 12v1 to 12v2 connection points.

It seem risky: I fear I could shortcircuit something.
BTW, I'm really a solder noob: in these years I managed to solder three tv remotes and a headphone volume control. Results: two remotes destroyed beyond repair, the thid is still waiting a led replacement and, incapable of repairing the headphone volume, eventualy I bypassed it. No thanks, me and a welder are the exact contrary of MacGyver with a paperclip and a few empty cans: he would make a tank, I would destroy it simply by touching

And before all that check if the rails are internally connected already with an ohm meter through the wires. [Many that say they are two rails on the output really aren't.]

Thank you, I'll check.

The Newton Power PSU's Ive seen were decent.

In fact I have one out of an old Gateway apart right now.
Checking if the caps in there might be the problem.
Answer: NOT.
Its full of Chemicon and Rubycon caps.
No crap caps at all.

Your words reassure me: in fact, I just want to avoid any psu failure for at least 5 years. If Newtons are sturdy but lack bell & whistles like a graphic card connector, modular cables or black satin case, they're welcome.
It's supposed to be a "mule" psu, not a "white horse" one. My maintenance should limit to a yearly cleaning from dust and [if necessary] a fan replacement.

The biggest problem with them is the OEM's tend to choose the smallest one that they can possibly get away with so you end up replacing them for even a little bit of an upgrade.

.

I suppose this is the reason they are sold on Ebay for a low price: according to the vendor, these are server psus the clients discarded because they wanted a bigger one or a redundant pair.
However, I'll ask if it's under warranty before its opening, just to be safe.

Zandrax

[PCBONEZ]

Re: New PSU for a dual 370

I don't see how you could possibly violate any regulation by have one rail vs two as long as you have the same number of wires.

Lets say
PSU1
12v1 - 5 wires passing 15 amps.
12v2 - 5 wires passing 5 amps

PSU2
One 12v - 10 wires passing 20 amps.

same same.

.

[gonzo0815]

Re: New PSU for a dual 370

I think FSP units are decent for this kind of system, at least from my point of view they are durable, even if not the best caps are used.

Sadly all the older Seasonic ones do not have good capacitors (e.g. SS400FS), otherwise they are very good (but bad capacitor at 5v Vsb can cause premature failure without preemptive recapping).

The problem with newer units is probably the shift toward 12v regulation, which can drive the 12v rail way out of the safe voltage range (e.g. FSP 350THN-P delivered 13v in a system based on Asrock K7s8XE+ & AMD Athlon XP 2000). due to group regulation.

The older Enermax EG365 where solid workhorses too, but recommendation is difficult due to unknown cap quality and issues with todays production (not sure if the older 5v regulated ones are affected).

[zandrax]

Re: New PSU for a dual 370

I don't see how you could possibly violate any regulation by have one rail vs two as long as you have the same number of wires.

Lets say
PSU1
12v1 - 5 wires passing 15 amps.
12v2 - 5 wires passing 5 amps

PSU2
One 12v - 10 wires passing 20 amps.

same same.

.

It's not a matter of wires, but of energy: EN 60950, section 1.2.8.8, states that a "hazardous energy level" is a 20 Joule energy or 240 VA for voltages above 2 v. So psu *must* limit their output to 240 VA on every line: while at 3.3 and 5v you can draw dozen of amperes, you aren't supposed to expect more than 20 A on the 12v line because an Over Current Protection (in short, OCP) should abruptly disconnect your psu from mains or severely [and riskly] reduce the voltage. So psu should have 2 or more 12 v rails, each one with an Ocp, in order to deliver the power required by today cpus and power-hungry, upper line gpus.

However, this recommendation isn't always respected: I remember a reviewer of a Seasonic unit, a couple of years ago, that managed to draw 30 A on the 12v rail: the Seasonic had absolutely no OCP on the 12v rails and, while this could be an advantage for gamers and overclockers, it could have block their sales in Europe. Instead of the Seasonic, I found a similar review of a Corsair HX620, which is Seasonc based.

Getting back to the Newton, I don't expect to draw, as stated on UL page, 19 A on every 12 v rail: it vould be absurd, as 19A * 12v * 2 rails = 456 VA which is greater than its 400W combined output. More likely, the 12v limit is around 20-22A, or 240-260 VA, which should be sufficient for the hard disks (P3 absorb mainly from 5 and 3.3 rail, so it should be quite safe).

Zandrax

[shadow]

Re: New PSU for a dual 370

I think FSP units are decent for this kind of system, at least from my point of view they are durable, even if not the best caps are used.

Yes I also agree that FSP power supplies are quite decent. An older model FSP would be quite good for the use that you envisage and it would not cost much either.

[PCBONEZ]

Re: New PSU for a dual 370

EXACTLY!

>>> limit their output to 240 VA on every line <<<

NOT on every 'rail'. ... On every *LINE*.

Rails have nothing to do with it.
By that thinking the 3.3v, 5v, and 12v would all count against each other.

You have many many 12v *lines* exiting the PSU.
And
You have the SAME number *lines* if it's one rail or two feeding them.
Same current in each of the *lines* whether feed by one rail or two.

[I am pulling this out of the "Way Back" Machine.]
Those lines happen to be 16 gauge wire which is limited to about 8 amps per wire.
On 12v more than about 96 VA in one wire and you'll melt the insulation and be able to light cigarettes.

You, we, they get around that problem having more wires (lines).
More "lines" are used for more current.
That's why there are so many wires for each voltage.
Thats why they added the 2x2 "P4" connector when P4 came out.
They needed more LINES to pass enough current through.

...

And, more wires ARE a matter of energy.
They use more wires to pass more energy at the same voltage.

.

[zandrax]

Re: New PSU for a dual 370

I think FSP units are decent for this kind of system, at least from my point of view they are durable, even if not the best caps are used.

Sadly all the older Seasonic ones do not have good capacitors (e.g. SS400FS), otherwise they are very good (but bad capacitor at 5v Vsb can cause premature failure without preemptive recapping).

The problem with newer units is probably the shift toward 12v regulation, which can drive the 12v rail way out of the safe voltage range (e.g. FSP 350THN-P delivered 13v in a system based on Asrock K7s8XE+ & AMD Athlon XP 2000). due to group regulation.

The older Enermax EG365 where solid workhorses too, but recommendation is difficult due to unknown cap quality and issues with todays production (not sure if the older 5v regulated ones are affected).

I agree: Fortron is in my second position and if I have told Newton units are prone to failure, I would buy a FSP400-60GLN, which is Atx 2.01 and still available. It's quite tiny, reasonably priced (45 € + shipping), Active Pfc, unfortunately not high 3.3 and 5v combined output (150W) and the airflow should cold their Teapos better than in the ATX-400PNF.

I wouldn't buy a second-hand psu until I can open it and check there isn't any blown or bad brand capacitor inside, because I can't replace them.

Zandrax

[zandrax]

Re: New PSU for a dual 370

EXACTLY!

>>> limit their output to 240 VA on every line <<<

NOT on every 'rail'. ... On every *LINE*.

Rails have nothing to do with it.
By that thinking the 3.3v, 5v, and 12v would all count against each other.

You have many many 12v *lines* exiting the PSU.
And
You have the SAME number *lines* if it's one rail or two feeding them.
Same current in each of the *lines* whether feed by one rail or two.

[I am pulling this out of the "Way Back" Machine.]
Those lines happen to be 16 gauge wire which is limited to about 8 amps per wire.
On 12v more than about 96 VA in one wire and you'll melt the insulation and be able to light cigarettes.

You, we, they get around that problem having more wires (lines).
More "lines" are used for more current.
That's why there are so many wires for each voltage.
Thats why they added the 2x2 "P4" connector when P4 came out.
They needed more LINES to pass enough current through.

...

And, more wires ARE a matter of energy.
They use more wires to pass more energy at the same voltage.

.

I've got the picture: "line" means the single cable between the psu and a connector, while I thought it was the effective output of a transformer. The output [say 12v] is routed into one or more rails [12v 1, 12v 2, ...], and each rail feeds some lines [brunch of cables].

However, it's unlikely that a single line can draw 20A: as you wrote, an Awg16 cable [quite thick for a standard psu, mine have Awg18 or 20, Allied excluded] can tolerate max 8 A, over that its insulation will melt. I hope yours are flameproof cable so they will simply melt, otherwise they will seriously burn and catch a fire.
Imo, the limit is on the single rail: it would be both dangerous and meaningless on a single line.

Zandrax

[PCBONEZ]

Re: New PSU for a dual 370

Yes that is how I interpret what 'line' means the way they are using it.

.

I wrote what I did because a single line can't pull more than 8 amps without melting.
That was my point. Maybe I worded it wrong.
I have some PSUs with AWG-16 and some with AWG-18.

This is right out of the ATX12V rev 2.2 Specification:

18 AWG is suggested for all wires except for the +3.3 V sense return wire, pin 13 (22 AWG).
For 300 W or more configurations, 16 AWG is recommended for all +12 VDC, +5 VDC, +3.3 VDC, and Common.

Note: It's only "recommended".

.

[Generally]
12v, 5v, 3.3v, -12v, and -5v (if present) all come off different secondary windings of the same transformer.

After that transformer there are individual voltage regulators for each 'rail'.
A PSU with two 12v rails internally would have two 12v voltage regulators but they would (probably) still be feed by the same transformer.

Now, if you tie the outputs of those two 12v regulators together you only have one rail *externally* but it is feed by two regulators in parallel.

This configuration has the advantage of splitting the load across the two regulators so that you avoid the situation where one regulator is heavily loaded and the other is barely loaded at all. Overall the 'system' is more stable.
-
However:
I just noticed the regulators in you pictured PSU have different capacities. (By the amps rating.) In that case bridging the two rails may not be a good idea because with non-matching components too much load may shift to the lighter duty regulator.

.

[davmax]

Re: New PSU for a dual 370

After that transformer there are individual voltage regulators for each 'rail'.
A PSU with two 12v rails internally would have two 12v voltage regulators but they would (probably) still be feed by the same transformer.

Now, if you tie the outputs of those two 12v regulators together you only have one rail *externally* but it is feed by two regulators in parallel.

Most PC power suplies do not have seperate regulators for +12V, -12V and 5V rails and 3.3V. Each voltage has a choke and rectifier. Each voltage is summed and fedback to the controller chip (3.3V excepted in example below). The transformer winding ratios play an important role in roughly setting the different supply rails. See example here:http://www.pavouk.org/hw/en_atxps.html

In the example the 3.3V circuit has a small stabilizing shunt regulator. With a 22 ohm shunt resistor this plays only a small role relative to the output current potential.

Too much heat would be generated having a seperate regulator for each rail at some very high currents.

[zandrax]

Re: New PSU for a dual 370

@ PCBONEZ:

Thank you for the explanation.
I think you are right about cable thickness: my psus are rated for 150 and 200 W, their cables are mainly Awg18. I'll check the Newton cables.

If I can choose the model, I'd pick the NPS-400AB B one, because its 12v rails are fairly balanced.


@ Davmax:

I have no idea: I've found at least four mosfets attached to heat sinks inside every psu, so I've thought each of them is a regulator for a voltage.
Since I'm not a skilled electrician and I can't read circuit schematics, I can believe you and think that some output are regulated by a shunt regulator [wikipedia page].
A doubt: the heat produced by some resistance is supposed to be greater than the one produced by a mosfet, isn't it?

Zandrax

[PCBONEZ]

Re: New PSU for a dual 370

davmax is correct but there are a lot of different kinds of power supplies.

I was referring to MOSFETs as voltage regulators.
They are actually only part of a complete regulator but they are the 'work' part and many people simply call them voltage regulators.
There is other circuitry (usually with a controller IC) that controls them turning on and off and that indeed would be part of the complete regulator.

I've attached a schematic from and old Deer 200 watt that shows what I meant about them all being supplied my the same xfmr ( T3 in the drawing. ) and it also shows what davmax is talking about.

.

[davmax]

Re: New PSU for a dual 370

I have no idea: I've found at least four mosfets attached to heat sinks inside every psu, so I've thought each of them is a regulator for a voltage.
Since I'm not a skilled electrician and I can't read circuit schematics, I can believe you and think that some output are regulated by a shunt regulator [wikipedia page].
A doubt: the heat produced by some resistance is supposed to be greater than the one produced by a mosfet, isn't it?

Zandrax

As PDBONEZ states PSU's vary, as do the number of MOSFETs. There is usually one MOSFET for the main switching regulator (all outputs) sometimes there is a preregulator using a MOSFET for active power factor correction. Other components on heat sinks can be rectifiers and can be mistaken for MOSFETs.
All regulation is switching technology to reduce heat loss/maintain efficiency.
The one regulator that controls all outputs takes a sample from several outputs to control the output voltages in a limited fashion. eg if both the +5 and +12 rails are sampled and a heavy load affects the +5 the output drop will be compensated upward however this will cause the +12V to rise and this will be compensated downward by the one regulator, the end result is a compromise thus the regulation of these supplies is not tight, the required tolerances allow for this. The main aim is power with efficiency.

[zandrax]

Re: New PSU for a dual 370

Update:

Yesterday the Newton arrived: since there isn't any "Void warranty if removed" seal, I opened it and took some photos.
It's better posting the photos here or in the "Power supply build quality pictorial" thread [the "Power supply build quality pictorial DISCUSSION" one doesn't seem in topic]?

Zandrax