Which PSU would you buy?

[kony]

More or less, but you have to integrate over the product
of the V and A waveforms.

BUT, in the context used, one cannot just multiply the rated
V and A and add them up to achieve a combined 3V+5V rating,
the result will be wildly inaccurate which is why I wrote
that one needs the specific parameter of combined 3V+5V
spec.

To put it concisely, the 3V rail is (most often) derived
from the 5V rail. There are rail weighting issues for
feedback purposes, such as how much 3V can be drawn before
it drops 5V rail too much, but more generally the total amps
the 5V rail can supply, alone, is all there is except that
it is derated by the amperage rating of the dual diode after
the transformer. That diode pair (or dual package) rating
can vary a lot depending on the quality of the PSU, and in
some passive designs they'll even parallel multiple diodes
to raise the rating (or reduce the derating by temp, however
you want to look at it). Then by the same token, 3V rail
also has a max current spec, and the two combined are a
specific spec provided by the manufacturer on many good PSU.

Some modern PSU may have relatively high wattage rating due
to high single or moderate dual 12V rails, but rather poor
combined 3V+5V rating like 130 or 150W. They are
inappropriate choices for an Athlon system above (roughly)
1.3GHz (more or less, depending on other component use of 5V
rail such as # of HDD, video card, etc).

To give examples, I have a Delta 300W PSU with a combined
3V+5V rating of 245W. These are conservative ratings too,
it can supply this wattage continuously. It would be (and
is, as currently employed for over 18 months if not longer)
quite capable of running an Athlon XP @ 2.5GHz using 5V
derived vcore power (CPU voltage regulation subcircuit
consuming over 100W by itself off of the 5V rail).

To contrast, you could instead buy a 620W Enermax, which
after derated based on ambient temp (per their specs) to
70%, will have 170W * 70% = 119W. For a $150 after rebate
PSU (
http://www.newegg.com/Product/Product.asp?Item=N82E16817194004
) this would be an incredibly bad choice for any Athlon
based system with a motherboard using 5V for CPU.

Note that many motherboards designed in, and originally
spec'd for, higher speed (and FSB) Athlon XP do use 12V
power for CPU instead of 5V. Either can work, and "if"
anyone decides they want to use a 5V based board to o'c one
to the levels I did, they might want to consider a lower
impedance path to the VRM subcircuit. In other words, to
overcome the board power plane or trace resistance, it can
help to add a low gauge jumper wire, it has a significant
measureable benefit in reducing voltage drop, for example of
one such modification to a Biostar M7NCG-400,
http://69.36.189.159/usr_1034/M7NCG_5V_Mod.jpg
I think I used 12 ga. single strand on that and it raised
voltage by about .3-.4V. YMMV on other boards but it always
seems to raise by at least 0.2V if not more.

The Enermax example above may be extreme in that it costs so
much for such a low combined and derated 3V+5V, but other
options aren't necessarily desirable either. Antec is
popular so let's look at their Neo HE 550W (currently $125
delivered at newegg,
http://www.newegg.com/Product/Product.asp?Item=N82E16817103941

http://www.antec.com/specs/NeoHE550_spe.html

It has a combined 3V+5V (taken liberally, since they don't
provide this combined spec anymore on this unit, so it is
optimistic, a best theoretical rating, of 179.2W.

Clearly if one wants a cost-effective PSU for a system using
higher current 5V based CPU power, choosing a quality older
ATX spec unit is beneficial. To brang a bit, the above
Delta with 245W combined 3V+5V power cost about $15 or $25,
I forget which as I bought volumes of them at two different
price-points. Either way, they are server-grade units,
there are no 300W units built as well even from PC Power &
Cooling, you'd have to buy a server or server PSU from
select manufacturers to get this kind of rating system...
today they would be called 400+W on the retail market.

However, there are plenty of other choices. Antec did make
a few older generations that had sufficiently high 3V+5V
ratings, but the fundamental shift that Intel pushed towards
12V current capability mean that current generation PSU
models just aren't cost effective for older platforms unless
at least new enough to use 12V power for CPU as typically
evidenced by the inclusion of the 4 pin 12V connector on the
motherboard. There are a few rare boards that can (do)
power 12V to the CPU VRM subcircuit by the lone 12V lead in
an ATX 20 pin connector, but I wouldn't want to own any of
them, let alone rely on one.

 

[Timothy Daniels]

Nope, no weave.. A specific context was written, I replied,
you disagreed, and now you think that if we disagree then I
will eventually try to turn around and agree with you in
some subtle way. I don't.

So clearly it is not anything like a weave, but now you
muddy the waters with silly attempts at a personal attack
instead of just sticking on topic.

If you try really really hard, possibly you can focus on
what WAS written instead of trying to make up something to
argue about, instead of trying to imply I meant more, or
other, than was written. If you GUESS I meant something I
didn't write, that's your own fault. If I then write more,
and it doesn't go along with what you had just GUESSED,
instead of what I wrote, your own mind has played a trick on
you but my stance on the topic remained static.

Would you like an example?

I recall you like PC Power & Cooling. OP wants a quiet PSU,
let's look at the specs for a Silencer 360, since it is near
the same price-point and should be adequate for the system.
Here's the spec sheet,
http://www.pcpowercooling.com/products/assets/S36ATX/specs.pdf

What do they spec? 1V p-p. When?
On a static load? No, it can easily do far better on a
static load, only load dynamics require such a high 1V spec.

How about efficiency? Did you think a PSU "must" score
around 74% if designed for a static load? Absolutely not.
That would be horrendous efficiency for a switcher designed
for a static load, but entirely plausable for one handling
dynamic loads.

Regulation, 5% and 10%. If they could reliably guarantee it
was always an order of magnitude less, they would do it.
Why can't they? It has to be spec'd for the intended use,
in a system that is a dynamic load.

Could someone hook it up to a system and perform one kind of
isolated test and find it less than 5 or 10% off? Of
course, and OTHER tests may deviate more... which is WHY
they don't try to lie and only mention best-case scenarios
instead of the truth.

Do you not realize that any bottom of the barrel generic can
legitimately claim less than 5 or 10% if it were only a
static load?

PC Power & Cooling makes some very good units, I'm not
trying to suggest otherwise but rather it was an example
from a company you seem to have liked, and they accept same
thing I wrote, that the dynamic loads DO result in worse
regulation on ALL of them.

That is just the tip of the iceburg though, the real issue
in many cases is the effect on the capacitors, the
performance of a part that has no need for bells, whistles,
or neon, has everything to do with whether it keeps working
instead of failing. It IS a performance parameter. One can
burn a candle at both ends but is it a high performance
candle or just one that burnt out fast? MTBF is yet another
PSU spec, they are all taken in context, not apart.

As I've written too many times to count, a short term test
is not a validating one, it only finds faults. Accelerated
wear is a poor performance index, and relative to a static
load, all PSU perform poorly.

Or to put it another way, we're contrasting static with
dynamic, so we're contrasting the same unit's accuracy when
static, vs. dynamically loaded. I did write RELATIVE
previously though, even capitalized and you still missed it.

Cut the long-winded crap. All the OP wanted was a
reasonable recommendation.

*TimDaniels*

 

[UCLAN]

Nope, no weave.. A specific context was written, I replied,
you disagreed, and now you think that if we disagree then I
will eventually try to turn around and agree with you in
some subtle way. I don't.

So clearly it is not anything like a weave, but now you
muddy the waters with silly attempts at a personal attack
instead of just sticking on topic.

If you try really really hard, possibly you can focus on
what WAS written instead of trying to make up something to
argue about, instead of trying to imply I meant more, or
other, than was written.

What *was* written was the question: Do any PC PSes perform well on
static tests but poorly on dynamic tests, other than the first versions
of the Antec-Seasonic Neo HE?

And your answer *was*: yes, they all do. [Note the word *all*.]

I am not guessing or implying anything. I am making up nothing. Your
answer was at best misleading, at worst plain WRONG. All attempts to
get you to qualify or explain your answer have resulted in your bobbing
and weaving. The thread is there for all to read.

 

[manny]

kony wrote:

Isn't it just W=V*A?

A person who can bench 300 lbs. and leg press 1000 lbs. usually
can't press and bench that much total weight simultaneously.

 

[kony]

Cut the long-winded crap. All the OP wanted was a
reasonable recommendation.

So what ends up happening?

OP asks "which PSU" as if this topic hasn't been covered
myriad times in this and other groups. There wasn't even
basic detail of the application and you can count on same
thing different day, over and over again. Better to just
get the idea out there that if they have a special, unusual
application then it requires more scrutiny but not just a
vanilla system with modest needs.

I suppose the reasonable recommendation would be "Google for
it since this has been covered, unless you have a special
need."

 

[kony]

What *was* written was the question: Do any PC PSes perform well on
static tests but poorly on dynamic tests, other than the first versions
of the Antec-Seasonic Neo HE?

And your answer *was*: yes, they all do. [Note the word *all*.]

I am not guessing or implying anything. I am making up nothing. Your
answer was at best misleading, at worst plain WRONG. All attempts to
get you to qualify or explain your answer have resulted in your bobbing
and weaving. The thread is there for all to read.

So we can concluded that your preconceived notions and
guessses about what I meant, mislead you.

Once that had happened, you feel your mistaken concept
wasn't the same as what I had meant, then it must be a bob
or weave to bring the two concepts together. Problem there
is that it wasn't a change in my posts but your mind.

 

[Timothy Daniels]

"koRny" whined:

So what ends up happening?

OP asks "which PSU" as if this topic hasn't been covered
myriad times in this and other groups. There wasn't even
basic detail of the application and you can count on same
thing different day, over and over again. Better to just
get the idea out there that if they have a special, unusual
application then it requires more scrutiny but not just a
vanilla system with modest needs.

I suppose the reasonable recommendation would be
"Google for it since this has been covered, unless you
have a special need."

You don't stop to THINK. If OPs keep asking the same
question, it's because they're not Googling. If you post
all your long-winded explanations, they *still* won't find it.
If you want to help, 1st ask the OP for info that you need
to respond reasonably. That in itself will be an education
enough, and those who do Google will find it. You can't
educate the world by boring people to death with lectures,
complaints about industry standards, and endless ponti-
fications.

*TimDaniels*

 

[Timothy Daniels]

So we can concluded that your preconceived notions and
guessses about what I meant, mislead you.

Once that had happened, you feel your mistaken concept
wasn't the same as what I had meant, then it must be a bob
or weave to bring the two concepts together. Problem there
is that it wasn't a change in my posts but your mind.

Man with keyboard diarhea not think before spewing crap.
Then must dodge and weave to avoid getting crap on face.

*TimDaniels*

 

[Timothy Daniels]

"Bob Day"

More or less, but you have to integrate over the product
of the V and A waveforms.

Hmmm.... As I dimly recall, that would be to integrate
the square of the product of their instantaneous values
over one cycle, divide by the interval, and then to take
the square root of the result - which would be called the
Root Mean Square (i.e. the square root of the average
value of the squares). I *think* that for periodic waveforms,
W(rms) = V(rms)*A(rms). And for direct current, that would
be simply W=V*A. Please correct me if I'm wrong.

But I think the point that koRny is *trying* to make is that
the max wattage value available at one voltage varies with
the wattage being output at the other voltage. IOW, they're
not independent, and their independence is a measure of
the PSU's quality.

*TimDaniels*
*TimDaniels*

 

[kony]

"koRny" whined:


You don't stop to THINK. If OPs keep asking the same
question, it's because they're not Googling. If you post
all your long-winded explanations, they *still* won't find it.

Won't find the simple answer, possibly. Might learn enough
to make their own decisions? More likely.

If you want to help, 1st ask the OP for info that you need
to respond reasonably.

Seems like I did mention lack of info more than once.

That in itself will be an education
enough, and those who do Google will find it. You can't
educate the world by boring people to death with lectures,
complaints about industry standards, and endless ponti-
fications.

Perhaps this is why some people still have to ask, because
they don't want to hear how to DIY. It is always amusing
though, that if I write more than a couple sentences it is a
pontification to you, but then if you do it, it must be some
grand insight into the origin of the universe. There's a
difference though, in that I make suggestions that tend to
counter possible problems, while you make suggestions where
there were no problems (like adding a fan to blow on eggs in
an incubator).

 

[kony]

Man with keyboard diarhea not think before spewing crap.
Then must dodge and weave to avoid getting crap on face.

*TimDaniels*

It's always nice to know what you're thinking about Tim.
Not everyone thinks of runny feces all day.

 

[kony]

But I think the point that koRny is *trying* to make is that
the max wattage value available at one voltage varies with
the wattage being output at the other voltage. IOW, they're
not independent, and their independence is a measure of
the PSU's quality.

Independence could be but is not necessarily a measure of
the PSU quality, but whether the design attempts to reduce
current bottlenecks (good parts spec margins) instead of
choosing a cheaper or smaller part.

In the end the issue is the same as always, having the spec,
the terms it was derived under (sustained, peak, temp, etc)
and the degree to which one can trust that spec. Having the
accurate information of the true capability one then has the
ability to match to the system needs. The combined
current rating could be completely dependant and it can
still be a quality design, it just needs to be spec'd such
that the information is applicable to determining if the
output is acceptible for the load.

 

[UCLAN]

What *was* written was the question: Do any PC PSes perform well on
static tests but poorly on dynamic tests, other than the first versions
of the Antec-Seasonic Neo HE?

And your answer *was*: yes, they all do. [Note the word *all*.]

I am not guessing or implying anything. I am making up nothing. Your
answer was at best misleading, at worst plain WRONG. All attempts to
get you to qualify or explain your answer have resulted in your bobbing
and weaving. The thread is there for all to read.

So we can concluded that your preconceived notions and
guessses about what I meant, mislead you.

"Can concluded" ?? What does that mean? The *fact* that the OP asked
if any PSUs act *poorly* dynamically, and you answered "all of them",
and later explain that most perform *worse* dynamically than statically.
Well, *poorly* and *worse* are two completely different things, so your
answer was quite misleading. You also have trouble confusing actual
performance with worse case specification. (If a unit has a regulation
specification of 1%, that's a maximum. Actual performance might be a
small fraction of that 1%.) PSUs are not specified that way for dynamic
loading, anyway. They are specified for maximum excursion, and time to
return to *static load* regulation.

Once that had happened, you feel your mistaken concept
wasn't the same as what I had meant, then it must be a bob
or weave to bring the two concepts together. Problem there
is that it wasn't a change in my posts but your mind.

Great rationalization. Hope you didn't hurt yourself writing it.

 

[Timothy Daniels]

"koRny squid" whined:

Perhaps this is why some people still have to ask, because
they don't want to hear how to DIY. It is always amusing
though, that if I write more than a couple sentences it is a
pontification to you, but then if you do it, it must be some
grand insight into the origin of the universe. There's a
difference though, in that I make suggestions that tend to
counter possible problems, while you make suggestions where
there were no problems (like adding a fan to blow on eggs in
an incubator).

Dodge and weave, dodge and weave. Change the subject,
modify and re-arrange history, hedge, avoid direct questions...
all your crap is in this thread. Dodge and weave and hide
behind a smokescreen of spurious verbiage. You are a squid.

*TimDaniels*

 

[kony]

... OP asked
if any PSUs act *poorly* dynamically, and you answered "all of them",
and later explain that most perform *worse* dynamically than statically.
Well, *poorly* and *worse* are two completely different things, so your
answer was quite misleading.

Ok, I suppose it could be misinterpreted, but that threshold
of poorly and worse is a fine line of distinction.
Personally, I'd consider a PSU swinging by 10% to be
undesirable, poor rather than only relatively worse.

You also have trouble confusing actual
performance with worse case specification. (If a unit has a regulation
specification of 1%, that's a maximum. Actual performance might be a
small fraction of that 1%.)

Yes it "might" be, or it "might" be worse, which is why they
have the spec to begin with. One cannot test under only one
load scenario and presume it'll always perform identically
under different scenarios, thus the whole reason why specs
exist and we haven't even considered larger sample sizes,
the variation possible from specimen to specimen.

PSUs are not specified that way for dynamic
loading, anyway. They are specified for maximum excursion, and time to
return to *static load* regulation.

They are spec'd (with any reputable company) for their
target use. They are not spec'd to power steady state
resistive loads, they are not marketed for powering your
garden lights, toaster oven, etc. They are spec'd for
systems known to be dynamic loads. The hint here is they're
"ATX" computer power supplies, a purpose-built device.

 

[kony]

"koRny squid" whined:


Dodge and weave, dodge and weave. Change the subject,
modify and re-arrange history, hedge, avoid direct questions...
all your crap is in this thread. Dodge and weave and hide
behind a smokescreen of spurious verbiage. You are a squid.

Tim, if I'm a squid, wouldn't that be an inkscreen instead
of a smokescreen?

LOL, you could always PLONK me if it it upsets you so much.

 

[Bob Day]

"Bob Day"


Hmmm.... As I dimly recall, that would be to integrate
the square of the product of their instantaneous values
over one cycle, divide by the interval, and then to take
the square root of the result - which would be called the
Root Mean Square (i.e. the square root of the average
value of the squares). I *think* that for periodic waveforms,
W(rms) = V(rms)*A(rms). And for direct current, that would
be simply W=V*A. Please correct me if I'm wrong.

You're wrong (mostly). And in so many ways that I'm not
going to bother correcting you. Refresh your knowledge on
this subject (Google is your friend) and then maybe we can
talk about it.

-- Bob Day

 

[Timothy Daniels]

You're wrong (mostly). And in so many ways that I'm not
going to bother correcting you.

I figured you couldn't.

*TimDaniels*

 

[UCLAN]

Ok, I suppose it could be misinterpreted, but that threshold
of poorly and worse is a fine line of distinction.
Personally, I'd consider a PSU swinging by 10% to be
undesirable, poor rather than only relatively worse.

Why don't you just suck it up and admit that you misread the
OP's question?

They are spec'd (with any reputable company) for their
target use. They are not spec'd to power steady state
resistive loads, they are not marketed for powering your
garden lights, toaster oven, etc. They are spec'd for
systems known to be dynamic loads. The hint here is they're
"ATX" computer power supplies, a purpose-built device.

Huh? Gobbly gook.

A typical dynamic loading spec is:

Load step: 25%
Maximum excursion: ___mv
Maximum time to return: ___uS

Some PSUs do fine. Others do not. For you to claim that they all do
"poorly" is irresponsible.

 

[Bob Day]

I figured you couldn't.

Well, you're wrong on that also. But I'm not going to bother
correcting you on that either. Bye.

-- Bob Day