(PeteCresswell) said:
That's my bias too. In my case it's bc I assume that I don't
know enough to make a very detailed judgment.
Here's what I gleaned from NewEgg:
========================================================================
- The power supply must output at least 18A (amps) on the +12Vrail(s)
for a mainstream up-to-date computer; more than 24A for a system with
a single enthusiast-class graphics card; and no less than 34A
when it comes to a high end SLI/CrossFire system.
You should *compute* the required power. Assume roughly 50W for
motherboard and RAM, coming from either 3.3V or 5V rails. I like
to see the rating "3.3V @ 20A" and "5V @ 20A" as minimum ratings
on the supply, to satisfy the lower rail ratings. The label on
the supply will have a "combined power rating in watts" for the
3.3V and 5V rails, and you would want that to be well
above the 50W number, so there is some headroom available.
Your storage devices need 5V, so that's one other form of loading
that comes out of that combined rating.
Disk drives take 12V @ 0.6A, and 5V @ 1A (gross estimate). CD/DVD/BD
drives can vary on boiler plate rating. I've measured 12V @ 1A on
a spinning CD drive. 12V @ 1.5A or 12V @ 2.5A is the boiler plate rating
on DVD or BD drives. The 5V rail on those might be 1.5A.
Processor watts, comes from the 12V rail. If a processor was 130W, you'd
divide by 12V to get amps. And scale the result, by dividing by
90% efficiency (0.9) on the VCore regulator, to get 12V2 current rating.
(130/12)/0.9 = 12A on 12V2.
Graphics cards, you look those up on Xbitlabs. Some cards draw a couple
hundred watts from the 12V rail. The exact value will be listed on Xbitlabs.
They've stopped measuring modern cards with their traditional method, so
for the newest of the video cards, we're kinda screwed now. When a card
is idle, the power is lower. But when planning a power supply purchase,
we assume the graphics card is "flat out" and drawing that couple hundred
watts. A low end video card, can be as low as 3.2 watts on the other hand,
and virtually "negligible" with respect to this power calculation.
It would be graphics cards that make a big difference, to your
purchase plans. Installing four graphics cards at 200W each,
gives you some idea how that can happen.
You total up your amps on each rail, you total up the watts consumed,
and that tells you what power supply you need.
And, if you think you're doing it right, consider that *every* number
printed on the PSU label, means something. If you felt there was
some number there that could be ignored, "you're holding it wrong"
- For residential and commercial users, only the active power is
measured and charged (apparent power is measured and charged for
industrial users)...
This is my take on PFC.
There are three types. None, passive PFC, active PFC. PFC controls
a type of power consumption, that as a home user, you're not typically
billed for. Only the "real" power component (in the vector sense),
appears on a home user power bill.
So whether a power supply has PFC is moot from that perspective.
If you work in the IT department of a major company, then yes,
you do care about billable imaginary power component. But not
as a home user (in North America).
What it does affect, is some capacity planning by the power company.
The power company thanks you, if you wasted additional money adding
PFC to your power supply.
Governments can mandate the addition of PFC to ATX supplies. The
power company thanks the nameless government official for this.
On the down side, PFC tends to interact with cheap UPS devices.
In previous years, a UPS could emit close to a square wave when
running from its battery. Later, they moved to a modified or
step sine wave. The active PFC circuitry hates those waveforms,
because the active PFC is stupid and assumes the AC voltage waveform
is the trusted sine wave, and it tries to make the amperes drawn,
fit the waveshape of the voltage waveform. If you feed it a square
wave on voltage, while the UPS runs on battery, the PFC tries to make a
square wave from the current draw. This is not a particularly
good thing. Even if it doesn't blow right away, it probably
isn't good for that little PFC circuit board.
To stop that, you mix Active PFC power supplies, with "pure sine"
UPS boxes. Or, alternately, just remove your UPS entirely, and
put up with the dirty file system that results from an uncontrolled
power outage.
So while "Active PFC" keeps the capacity planning manager at
your power company very happy, it can involve extra expense
if you want to "do it right" with respect to any UPS you have
installed as well. It's one reason, when I needed a replacement
supply for one of my older PCs, I used a non-PFC supply for that.
Because I could be assured my crusty old UPS (on its second battery),
wouldn't blow the thing up at some point.
If it wasn't for the potential interaction with the UPS I bought,
I'd be all for active PFC. If your PC plugs straight into the
wall plug, active PFC is a good thing. (The power company thanks
me for this endorsement.)
A power supply with a higher efficiency rating will not only
help save costs - the heat dissipated will be much lower as well,
Higher efficiency supplies are OK, up to a point. The very highest
efficiency kind, they use two stage regulation. The main section
does 12V output at some high max current. The 3.3V and 5V rails
consist of a DC-DC converter, which accepts 12V in, and makes the
3.3V and 5V the chipset on the motherboard needs. The problem with
this, is the DC-DC converter typically doesn't support high currents,
and if you're running an older PC which consumes lots of 5V,
that is exactly the wrong kind of power supply to own. (It means
that 87% efficient supply that costs $200, might not be very strong
on the 3.3V and 5V rails.)
So, say you saw an 87% efficient supply, an 83% efficient supply,
and a 65% efficient supply. If you had a modern PC (with that i7
processor in it), you might grab the 87% one. (Knowing that modern
PCs have pretty low 3.3V and 5V loadings.) If you were doing
a supply replacement for your old Athlon machine, which has no
12V cable in it, you'd probably examine the label carefully on
the 83% one, or go with the 65% efficient one which advertised
"5V @ 40A". In fact, there's no way to actually, safely, draw
the 40A, but the architecture of the supply is such, that it's
better able to provide the 5V the board uses. My old Athlon
board without a 12V cable, draws around 25A from the 5V rail, max.
So, most of the time, the 87% one is a good buy, good for
the environment, helps keep your power bills low. But if
you're repairing PCs for a living, it's not always the
correct choice. For crusty old PCs, a crusty old ATX PSU
is what you install with it. Not a new, fancy one.
- Noise is always a problem wherever fans involved.
The fan speed is seldom chosen with any common sense at all.
It seems to be a pretty arbitrary decision by the manufacturer.
The inside of the power supply could be burning hot, or
cool as a cucumber, and it's just arbitrary on their part.
It's like they're not even thinking about it very much.
Only when the PSU shuts down on you, do you get an inkling
what they've done, and how stupid it was.
Some supplies keep the fan "cranked", and those are likely
the supplies rated for operation in a 40C room temp.
- There are also some "fanless" power supply products that
utilize passive cooling and are without any fans and other
moving parts.
Fanless supplies are perfectly acceptable. Especially if
the PC wasn't a barn-burner in the first place. If you have
a low end dual core, are using the integrated graphics (no
added video card), one of those could be fine. Do the calc.
for yourself, and see what you need.
Some fanless supplies, are relying on the computer case
additional fan for air movement, for some cooling. Others,
may actually work by making no assumptions about other
fans. Read the reviews, and see how existing customers
find them, whether their rating was honest or not.
Some of the fanless ones, make "coil noise". The thing is,
if you attempt to make a quiet PC, all that does, is start
to expose all the noises you couldn't hear before, because
the noises were drowned out by the sound of fans. Once you
remove the fans, you're still going to find noise sources.
- A power supply with modular cabling ability can reduce the
number of unused power connectors/cables in your system case,
and generally promotes tidier
and cleaner looking case internals as well as better air flow.
As long as the modular cabling doesn't cause too much voltage
drop, I'm all for it. Older modular designs, the cables
could go into the wrong connectors, with disastrous results.
Newer designs have learned from that kind of thing, and it
is less likely to be an issue now.
Paul
