Single-rail vs. multi-rail power supplies?

[Yousuf Khan]

I had the same thing and cured it by raising the CPU voltage slightly.

Charlie

Were you overclocking? I'm not overclocking, at the moment.

Yousuf Khan

 

[Loren Pechtel]

I actually did get one at one time, but it turned out to be bad, so I
returned it, and I never bothered again.

Yousuf Khan

Ours gets dragged out on occasion. It's convinced my wife that she
was wrong about the cost of various things.

 

[Arno]

I currently have a desktop with a Zalman 600W PS, it was a pretty nice
PS a few years ago when I got it, but now it looks like the system has
been upgraded and grown again, and I'm starting to see signs that it's
not producing enough power for the components anymore. So I'm looking at
the market for bigger PS's, likely 750W+.

When I got the Zalman, multi-rail +12V were the de riguer feature of the
day. Now that I'm shopping for them again, I see that the manufacturers
have done an about-face, and they are advertising the advantages of
single-rail +12V. The old 600W had 4 rails at 16 Amps each. The new
single rails I'm seeing seem to have a rail at anywhere from 45A to 65A!
Why have the manufacturers done the about-face on single-rail vs.
multi-rail?

I suspect marketing. The onld ones were not really muti-rail
either, more like "multi-rectifier". For proper multi-rail
you have to have a regulator, power switches, rectifier and
power-transformer for each rail.

It is also possible that recitifers have gotten better.

Arno

 

[Charlie]

Were you overclocking? I'm not overclocking, at the moment.

Yousuf Khan

No, but I had lowered the voltage to reduce the heat. Its an I7 CPU.
That didn't help much so I got a different heatsink which did help. When
I looked at the Event logs I noticed that the optical drive errors
started at about the time I lowered the voltage, so I raised the voltage
and haven't had an optical drive error since (about 1 year).

Charlie

 

[Arno]

No, but I had lowered the voltage to reduce the heat. Its an I7 CPU.
That didn't help much so I got a different heatsink which did help. When
I looked at the Event logs I noticed that the optical drive errors
started at about the time I lowered the voltage, so I raised the voltage
and haven't had an optical drive error since (about 1 year).

Overclocking has about the same effect as lowering the voltage,
signals may not get quite enough time. Overclocking is worse though,
as things can overheat in addition.

Arno

 

[Yousuf Khan]

No, but I had lowered the voltage to reduce the heat. Its an I7 CPU.
That didn't help much so I got a different heatsink which did help. When
I looked at the Event logs I noticed that the optical drive errors
started at about the time I lowered the voltage, so I raised the voltage
and haven't had an optical drive error since (about 1 year).

Well, then it seems like you put the CPU back to its recommended
settings. My CPU is already at its recommended settings.

Yousuf Khan

 

[Yousuf Khan]

I suspect marketing. The onld ones were not really muti-rail
either, more like "multi-rectifier". For proper multi-rail
you have to have a regulator, power switches, rectifier and
power-transformer for each rail.

Probably, but the effect was the same, namely that each of the 4 +12V
channels were limited in the amount of current they carried.

It is also possible that recitifers have gotten better.

Yeah, it had better be, or else I should be smelling smoke quite quickly.

Yousuf Khan

 

[Paul]

Probably, but the effect was the same, namely that each of the 4 +12V
channels were limited in the amount of current they carried.


Yeah, it had better be, or else I should be smelling smoke quite quickly.

Yousuf Khan

The only way a rectifier gets better, is by disappearing.
In the picture here, the rectifier on the secondary is
replaced by a MOSFET.

("Synchronous rectification")

http://www.onsemi.com/pub_link/Collateral/AND8460-D.PDF

Paul

 

[Charlie]

Well, then it seems like you put the CPU back to its recommended
settings. My CPU is already at its recommended settings.

Yousuf Khan

I did not raise the voltage back to where it was. Just enough to stop
the errors. I'm not saying that your errors are caused by the same
thing as mine. Just giving you something else to try.

Charlie

 

[Yousuf Khan]

The only way a rectifier gets better, is by disappearing.
In the picture here, the rectifier on the secondary is
replaced by a MOSFET.

("Synchronous rectification")

http://www.onsemi.com/pub_link/Collateral/AND8460-D.PDF

Interesting.

Yousuf Khan

 

[larry moe 'n curly]

I currently have a desktop with a Zalman 600W PS, it was a pretty
nice PS a few years ago when I got it, but now it looks like the system
has been upgraded and grown again, and I'm starting to see signs that
it's not producing enough power for the components anymore.

mainly the optical drives
sending controller error messages even when they are not being used, and
I'm also noticing some occasional spin retry errors on a few of my
internal HDD's.

Bit-Tech.net says the 1GB Radeon 6870 uses from 128W - 247W. Power
estimators aren't very good and usually give numbers that are way too
high, and when XbitLabs.com took actual measurements, they got this
(associated article goes into much more detail):

www.xbitlabs.com/images/coolers/system-wattage/p40.png

Or you could plug your system into a Kill-A-Watt and multiply the watt
rating by about 80% to get the needed PSU wattage. BTW the maximum
capacity of a PSU varies with temperature, and many years ago PC Power
& Cooling said an Enermax rated for 550W @ 25C (room temperature)
dropped to 366W @ 40C (not much hotter than some case interiors) and
244W @ 50C.

If the drives are PATA instead of SATA, have you tried tightening the
power connectors from the PSU? The connectors are metal tubes, and
prying between them and their surrounding plastic can make them fit
more snugly.

I've had optical drives based on Philips designs (Acer/Benq/Aopen)
suddenly act up and make head seeking noises, even when empty. I
could never figure out what went wrong (reseated internal connectors,
resoldered power chips, tapped all over the circuit boards, tried
freeze spray), but if the drives were removed for a few months they'd
work again, for up to a year.

 

[Rod Speed]

larry moe 'n curly wrote

Yousuf Khan wrote

Bit-Tech.net says the 1GB Radeon 6870 uses from 128W - 247W.

The datasheet says nothing like that.

Power estimators aren't very good and usually give numbers that
are way too high, and when XbitLabs.com took actual measurements,
they got this (associated article goes into much more detail):

www.xbitlabs.com/images/coolers/system-wattage/p40.png

And you dont have to measure with the 12V rail currents, they are in the datasheets.

Or you could plug your system into a Kill-A-Watt and multiply
the watt rating by about 80% to get the needed PSU wattage.

Makes a hell of a lot more sense to calculate the 12V rail currents using the datasheets.

And measure the system performance when you have a power supply
that you are wondering if its adequate using a decent max min multimeter,

BTW the maximum capacity of a PSU varies with temperature,

Its normally determined by the currents that protection circuitry shut down at.

and many years ago PC Power & Cooling said an Enermax rated
for 550W @ 25C (room temperature) dropped to 366W @ 40C
(not much hotter than some case interiors) and 244W @ 50C.

Dont believe those numbers.

If the drives are PATA instead of SATA, have you tried
tightening the power connectors from the PSU? The
connectors are metal tubes, and prying between them and
their surrounding plastic can make them fit more snugly.

Its better to release the contact from the shell and sqeeze the contact in a bit.

Or just try a different connector and see if the errors are seen with that one too.

In his case, because he has so many hard drives, it may even be the power Y adapters.

 

[Arno]

I did not raise the voltage back to where it was. Just enough to stop
the errors. I'm not saying that your errors are caused by the same
thing as mine. Just giving you something else to try.

In this case you will likely stikk have errors on occasion.
You need to raise it to where the errors apprarently stop, and
then a bit more. Of coruse if the errors are non-critical,
you do what you did. But you should be aware that this
can also cause other errors that may not be as visible.

Arno

 

[Arno]

The only way a rectifier gets better, is by disappearing.

No, definitely not. While syncronous rectifiers have their
merits, they are mostly for switchers without transformers,
as otherwise they drive cost up significantly. If you have
very low voltages (CPU, RAM), they are the only sensible
choice.

In PSUs with transformers, typically you use normal power
rectifiers, or shottky rectifiers. These have gotten slowly
better and cheaper.

Arno

 

[Arno]

Interesting.

Standard concept, but too expensive with transformers.

Arno

 

[larry moe 'n curly]

Standard concept, but too expensive with transformers.

My only PSU with a MOSFET rectifier in it was so cheaply built that PC
Power & Cooling featured it in an ad as an example of a bad PSU.

 

[Charlie]

In this case you will likely stikk have errors on occasion.

Well, I consider a year without an optical drive error to be sufficient
to believe the problem is fixed.

You need to raise it to where the errors apprarently stop,

I did.

and
then a bit more.

I can't see the point to that, but if I did start to get some errors of
unknown origin I would definitely raise the voltage to see if it would help.

Of coruse if the errors are non-critical,
you do what you did. But you should be aware that this
can also cause other errors that may not be as visible.

I'm having no other problems.

Charlie

 

[Paul]

Standard concept, but too expensive with transformers.

Arno

Um, yeah, that's the transformer the ATX supply needs anyway.

Compare the picture in AND8460-D.PDF to Pavouk, for some inspiration.

Transformer T3 in the upper right of this picture, is the equivalent
of the transformer present in the Onsemi appnote. The synchronous
rectifier idea, would replace the diode pairs on the right of T3.

http://www.pavouk.org/hw/en_atxps.html

The advantage of the MOSFET, is no longer having to deal with the
diode equation, and a rising voltage drop across the rectifier,
as the current flow through it rises. If I run 20 amps through
a rectifier diode, and the drop across it is 1 volt, that's 20W
of heat. You then compare that to the channel resistance of the
MOSFET (and the I**2*R loss there), and the amount of energy needed
to drive the gate capacitance (it's a factor too). The Onsemi appnote
shows 91% efficiency at mid-load (Figure 50 on PDF page 26).

Paul

 

[Yousuf Khan]

Bit-Tech.net says the 1GB Radeon 6870 uses from 128W - 247W. Power
estimators aren't very good and usually give numbers that are way too
high, and when XbitLabs.com took actual measurements, they got this
(associated article goes into much more detail):

www.xbitlabs.com/images/coolers/system-wattage/p40.png

I don't think the video is at issue at all, it's plugged into its own
private rail, and I've never had any problems with it. It's the storage
units that are having the problems.

Yousuf Khan

 

[Yousuf Khan]

I did not raise the voltage back to where it was. Just enough to stop
the errors. I'm not saying that your errors are caused by the same thing
as mine. Just giving you something else to try.

Well, I'll keep it in mind then. When all else fails, that's worth a
shot. But of course my first solution is a new PS.

Yousuf Khan