Cheap/generic PSUs - any good? (was: Low power socket A)

[David Maynard]

More likely the wattage ratings on them were way inflated,

Would seem likely.

ever seen those
"600 WATT" pc speakers with the 0.75W amp chip in them?

Hehe. Oh sure. Makes ya wonder why anyone bothers with those 2 grand home
stereo systems when a 6 buck pair of PC speakers outdoes them by 300 watts,
eh? <chuckle>

 

[kony]

I hate these extra worries....AMD can give big headaches, because of the
power requirments....

Compared to what, a Via C3?
It's easier, on average, to run an Athlon box than a P4,
excepting rare situations where a semi-modern Athlon XP used
5V for CPU and the video card ONLY used 5V->3.3/1.5V, no aux
connector like the more modern vidcards have. That's not
really a problem though, it simply means choosing a PSU with
~ 200W+ 3V+5V combined rating. On the other hand, similar
problem could've been found running a P4 with a lot of HDDs
on a marginal PSU

Otherwise, looking at both Intel & AMD's current offerings,
Intel's use more power, but both use 12V rail for CPU
(except a dwindling number of now-aging Athlon XP boards)
everthing else being equal the power supply need be
marginally higher capacity for Intel platform.

 

[larrymoencurly]

In my opinion, you get what you pay for. If you buy $2,000.00
plus of computer, and then throw in a $30.00 PSU, you are asking
for trouble.

I haven't paid more than about $15-20 for any PSU in the past couple
of years, but I've never gotten junk, except for a free-after-rebate
PSU + case. People can buy good stuff cheap from the surplus market
or by choosing Fortron/Sparkle/Hi-Q/PowerQ brand (Newegg has a 350W
for $32, delivered). I'm sure that these are better than other brand
$75-100 PSUs that come in fancy colors and include glowing fans.

 

[jakdedert]

I expect that the power ratings for the individual voltage outputs were
spec'ed in the same dishonest way that hifi manufacturers used to rate power
amps (still do, in fact). Each *individual* output would supply the
specified current or wattage; but load them all up at the same time and POW!

Amp manufacturers used to spec their output power one channel at a time, for
a *very* short time (even down to milliseconds). They could 'honestly'
state an inflated value, but real-world ratings were considerably less.

OTOH, I expect that there are seldom real-world operations on a computer
where 'all' of the given outputs on the psu are loaded to the max for any
length of time...thus allowing the cheaper manufacturers to get away with
the deception.

jak

 

[David Maynard]

I expect that the power ratings for the individual voltage outputs were
spec'ed in the same dishonest way that hifi manufacturers used to rate power
amps (still do, in fact). Each *individual* output would supply the
specified current or wattage; but load them all up at the same time and POW!

Well, an audio amp shouldn't 'pow'. What should happen with a single
channel power spec is you run into distortion before you reach what you
'think', from the channel spec, is max power output. (spec'ing single
channel power is done to cover the power supply's inability to provide that
much power to both channels simultaneously. Voltage output droops and
causes distortion)

Amp manufacturers used to spec their output power one channel at a time, for
a *very* short time (even down to milliseconds). They could 'honestly'
state an inflated value, but real-world ratings were considerably less.

Well, I don't think whether it's both channels driven or the bogus 'peak
music power' really fall into the same category. A single channel spec does
have some sense to it. 'Peak Music Power' doesn't.

But, for a completely 'honest' number one should be looking to continuous
RMS output, both channels driven.

OTOH, I expect that there are seldom real-world operations on a computer
where 'all' of the given outputs on the psu are loaded to the max for any
length of time...thus allowing the cheaper manufacturers to get away with
the deception.

Yes, that's often used as an excuse: won't happen in the 'real' world.

 

[Watson A.Name - \Watt Sun, the Dark Remover\]

jakdedert wrote: [snip]

OTOH, I expect that there are seldom real-world operations on a computer
where 'all' of the given outputs on the psu are loaded to the max for any
length of time...thus allowing the cheaper manufacturers to get away with
the deception.

Yes, that's often used as an excuse: won't happen in the 'real' world.

But you have to look at it from the other perspective. A PS might be
rated for one output at 30A and another at 20A (just for example). And
the _peak_ loads may be such that the 20A output is putting out 24A for
short times, while the 30A output is loafing at never more than 20A.
And on the average, the total wattage adds up to less than the rated
wattage. But because one PS was conservatively rated, that 4A over the
max doesn't harm it. But another PS wasn't so conservative, and the 4A
overcurrent causes it to fail, even tho the total wattage was within its
max.

So I think that the individual outputs should each be viewed as a
separate rating, even tho that complicates things. They should all
standardize on a format like giving 3 or 4 currents, for 5V, 12V, 3.3V,
etc. So you'd see 5V@30A/12V@15A/etc. Total wattage, like "350W",
isn't enough info to make an informed decision on the PS's capabilities.

 

[David Maynard]

jakdedert wrote:
[snip]

OTOH, I expect that there are seldom real-world operations on a

computer

where 'all' of the given outputs on the psu are loaded to the max

for any

with


Yes, that's often used as an excuse: won't happen in the 'real' world.

But you have to look at it from the other perspective. A PS might be
rated for one output at 30A and another at 20A (just for example). And
the _peak_ loads may be such that the 20A output is putting out 24A for
short times, while the 30A output is loafing at never more than 20A.
And on the average, the total wattage adds up to less than the rated
wattage. But because one PS was conservatively rated, that 4A over the
max doesn't harm it. But another PS wasn't so conservative, and the 4A
overcurrent causes it to fail, even tho the total wattage was within its
max.

So I think that the individual outputs should each be viewed as a
separate rating, even tho that complicates things. They should all
standardize on a format like giving 3 or 4 currents, for 5V, 12V, 3.3V,
etc. So you'd see 5V@30A/12V@15A/etc. Total wattage, like "350W",
isn't enough info to make an informed decision on the PS's capabilities.

Proper power supplies ***DO*** give the amperage per rail, and 3.3v & 5v
combined power, in addition to the 'total power'.

And if the ones you're looking at don't then don't go near them with a 10
foot pole,... or an 11 foot Hungarian either

Problem is, even when they do some of the el-cheapos lie about it, even on
the sticker.

What's even more incredible, some of them actually TELL you, right on the
sticker, that they're lying. It'll say something akin to "300W... actual
power 200W." Of course, the 300W is in the big letters and what is advertised.

 

[JAD]

oh BS MAYBE the prescott.............................. amd still has to
post their recommanded PSU's. I never even think a moment about PSU's
(within reason) when I put a P4 together.
Hey AMD stock is up relax the high pressure sales will ya

 

[kony]

oh BS MAYBE the prescott.............................. amd still has to
post their recommanded PSU's. I never even think a moment about PSU's
(within reason) when I put a P4 together.
Hey AMD stock is up relax the high pressure sales will ya

Maybe you should start thinking about it instead of making
assumptions. Read some spec sheets, knowledge beats
guesses any day of the week.

 

[Mike]

Without a load the PSU can't regulate (an inherent characteristic of
switching power supplies) but, while I can't say I've seen 'every' power
supply out there, I've never seen one 'burn up' from being unloaded.

Just the one: Supplied in error in place of the model that should have
been sent. The ordered model had no minimum current (internal ballast
resistors). The supplied model ... didn't.

Switch on ... tick tick tick stop ....

Died taking out numerous components. Returned for replacement.

This was not a PC supply though, just a generic Switch Mode.

The other is over-voltage protection (OVP). That circuit should clamp the
output to a 'safe' level if something in the PSU fails causing it to put
out excessive voltage. That would only matter if the fault specifically
caused an over-voltage rather than, say, the outputs going dead to begin
with. It's effectiveness is complicated by how fast it responds and how it
controls the voltage.

I think this was the cause of the failure. No current drawn, no ability
to sense output voltage. Voltage rises in an attempt to hit nominal
voltage, and (unfortunately) overshoots it. Protection kicks in and
clamps power supply output. Repeat until failure.

 

[kony]

On Sun, 26 Sep 2004 15:36:35 +0000 (UTC),

Just the one: Supplied in error in place of the model that should have
been sent. The ordered model had no minimum current (internal ballast
resistors). The supplied model ... didn't.

Switch on ... tick tick tick stop ....

Died taking out numerous components. Returned for replacement.

This was not a PC supply though, just a generic Switch Mode.

Yes, it is not uncommon for small, non-PC switchers to lack
(or at least, insufficient) integral loading.

 

[Watson A.Name - \Watt Sun, the Dark Remover\]

On Sun, 26 Sep 2004 15:36:35 +0000 (UTC),



Yes, it is not uncommon for small, non-PC switchers to lack
(or at least, insufficient) integral loading.

I'm thinking about all those laptops out there, that get used part of
the day and then get locked in a cabinet the rest of the time. All
those laptop chargers get plugged into the power when they're first
unpacked, and they're left plugged in 24/7, even tho the laptop is in
the cabinet. So most of the time these switchers are connected to the
AC line without any load.

 

[David Maynard]

Just the one: Supplied in error in place of the model that should have
been sent. The ordered model had no minimum current (internal ballast
resistors). The supplied model ... didn't.

Switch on ... tick tick tick stop ....

Died taking out numerous components. Returned for replacement.

No offense but your analysis of what caused the failure is, shall we say, a
bit abbreviated. One thing that sticks out is how did it 'take out numerous
components' if there was no load on it?

This was not a PC supply though, just a generic Switch Mode.




I think this was the cause of the failure. No current drawn, no ability
to sense output voltage.

It isn't that it can't *sense* the output; it can't regulate it.

A switcher operates with some form of pulse train 'switching' into the
output caps. Might be a constant frequency PWM (common) but the details
don't really matter. What matters is there is a minimum pulse width, or
minimum frequency, that the circuit can operate at and if that 'minimum'
output is more than needed to provide the (non existent) load current then
the sense circuitry is going to try to drive it even lower but it can't be
driven lower than minimum. So you're going to have either an overvolt
condition or, the more likely, switcher shutdown since 'less than minimum'
is PWM 'off'.

Voltage rises in an attempt to hit nominal
voltage, and (unfortunately) overshoots it. Protection kicks in and
clamps power supply output. Repeat until failure.

Once the output is clamped the story is over: it's shutdown; no
'on-off-on-off-on-off...'. OCP doesn't release till input power is removed.

Of course, that presumes things are designed properly.

 

[JAD]

first of all, these temp claims are soo clouded and suspect due to
power usage difference its hard to find a comprehensive chart without
a bunch of disclaimers and footnotes. buried in that techno babble is
the truth, but all in all the differences are minimal AND I never had
to buy 3rd party coolers for any Intel chip. Never burned one
either....which brings the 'throttling' thing in ...more 'issues'
which makes the facts corruptible.

 

[kony]

I'm thinking about all those laptops out there, that get used part of
the day and then get locked in a cabinet the rest of the time. All
those laptop chargers get plugged into the power when they're first
unpacked, and they're left plugged in 24/7, even tho the laptop is in
the cabinet. So most of the time these switchers are connected to the
AC line without any load.

Generally a "user detachable" power supply will have at
least minimal load integral, while other small switchers may
never be expected to be disconnected from rest of device, so
rest of device is always providing needed load.

 

[kony]

first of all, these temp claims are soo clouded and suspect due to
power usage difference its hard to find a comprehensive chart without
a bunch of disclaimers and footnotes. buried in that techno babble is
the truth, but all in all the differences are minimal AND I never had
to buy 3rd party coolers for any Intel chip. Never burned one
either....which brings the 'throttling' thing in ...more 'issues'
which makes the facts corruptible.

There is no cloud. Technical documentation and tests of
current usage on the opposing platforms both indicate that
the current-gen Intel chips use more power. As for past
CPUs, same avenues apply, the tech specs and field tests
reveal the truth.

 

[Mike]

Died taking out numerous components. Returned for replacement.

No offense but your analysis of what caused the failure is, shall we say, a
bit abbreviated. One thing that sticks out is how did it 'take out numerous
components' if there was no load on it?[/QUOTE]

It is abbreviated, it took a while to figure out what components had failed
etc. and then tell the supplier about it.

The fault took out numerous components because that is in the nature of
some SMPS designs when there is no load, and that's why it was returned. If
it had been a simple blown fuse, then it'd be a patch it and use it fix.
As it was, there were a number of semiconductors failed, looking at the
replacement cost (plus risk of repeat performance failure), it was easier
to just return it.

Once the output is clamped the story is over: it's shutdown; no
'on-off-on-off-on-off...'. OCP doesn't release till input power is removed.

Depends on design, surely? I've seen other PSUs keep tripping and restarting
when overloaded or "under" loaded. Not all of them fail though

Of course, that presumes things are designed properly.

That's a big presume there

 

[John Nagle]

Yes. Many cheap PC power supplies do not have UL certification,
and in some cases, the UL certifications are forged. UL certification
indicates that if loaded up to to its rated load, it will work, and
that if shorted or any single component fails, it will not catch fire.
There are many cheap power supplies out there which will NOT survive
a full load. Some will catch fire.

You can check UL certifications by going to "http//www.ul.com".
Find the UL file number (which will look like "E12345") on the
device and type it in. Or call the UL hotline at 888-854-6275.

Anything made in China should have a UL hologram label.
If it is from China, current production, and has a
non-hologram UL label, it is a counterfeit product.

Counterfeit labels should be reported to UL. See

http://www.ul.com/about/otm/otmv10n1/counterfeit.html

John Nagle

 

[David Maynard]

No offense but your analysis of what caused the failure is, shall we say, a
bit abbreviated. One thing that sticks out is how did it 'take out numerous
components' if there was no load on it?

It is abbreviated, it took a while to figure out what components had failed
etc. and then tell the supplier about it.

The fault took out numerous components because that is in the nature of
some SMPS designs when there is no load, and that's why it was returned.[/QUOTE]

Sorry, but I still don't know what that means: "the nature of?"

You mean components in the PSU? not other components it was (not) connected to.

If
it had been a simple blown fuse, then it'd be a patch it and use it fix.
As it was, there were a number of semiconductors failed, looking at the
replacement cost (plus risk of repeat performance failure), it was easier
to just return it.




Depends on design, surely?

It is, of course, always possible to do it wrong but, to use your term, it
is 'in the nature of' <g> the situation itself. Consider designing one that
'resets' automatically. It would ALWAYS oscillate (on a fault) since the
clamp, by definition, will put the output back into the 'safe' area causing
removal of the clamp which will then let the output go back to fault which
will cause a trip which...

I can't imagine why anyone would think that is 'desirable' operation (it
isn't as if this is a pole power line where we're burning off tree branches
with breaker reclosures) and, with that as (not) a criteria, there's just
no 'right way' to make one that resets as a result of it's own clamp action.

I've seen other PSUs keep tripping and restarting
when overloaded or "under" loaded.

Sounds like a logic error in the design then, or some odd and unforeseen
anomaly was causing it.

Not all of them fail though




That's a big presume there

Hehe. Yes. But predicting what a, uh, 'what is it'? circuit will do is an
exercise in voo-doo

 

[Tony]

Sorry, but I still don't know what that means: "the nature of?"

You mean components in the PSU? not other components it was (not) connected to.


It is, of course, always possible to do it wrong but, to use your term, it
is 'in the nature of' <g> the situation itself. Consider designing one that
'resets' automatically. It would ALWAYS oscillate (on a fault) since the
clamp, by definition, will put the output back into the 'safe' area causing
removal of the clamp which will then let the output go back to fault which
will cause a trip which...

I can't imagine why anyone would think that is 'desirable' operation (it
isn't as if this is a pole power line where we're burning off tree branches
with breaker reclosures) and, with that as (not) a criteria, there's just
no 'right way' to make one that resets as a result of it's own clamp action.


Sounds like a logic error in the design then, or some odd and unforeseen
anomaly was causing it.

Every packaged PSU module we have used has worked that way - Our
product goes into industrial sites, and it is an unfortunate fact of
life that accidental human-induced faults do happen occasionally, and
in those cases we do want the SMPS to keep trying to bring up the
line, as long as there is no damage. If it needed power removed before
power would be restored, THAT would be a "logic error" and a major
PITA for our customers.

Tony (remove the "_" to reply by email)