# How much power does a computer actually draw?

#### Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

•  Subject
• Author
• Posted on
Netters,

I am running four computers in a small network at home. All the four
mini-tower boxes are sitting side-by-site. There is only one monitor
that is shared using a KVM switch.

I am looking at getting a UPS system that must be adequate for these 4
computers. However, I am confused with various numbers that are thrown
around.

1. On trip-lite UPS website calculator, it claims that it machine must
take 77 watts on an average. This assumes each computer has its own
monitor.

2. Another website claims that the consumption should be around 50
watts per computer.

3. There is typically a power loss of 20% in the power supply.
Therefore, the actual power drawn is 77/0.8 = about 100VA.

4. I saw one more website that had a VA graph for a P4 computer+5
drives. This graph indicates an initial surge of 200VA (when the
computer was turned on) but an average of 170VA.

Questions:

1. So many different results for power consumption. What do you think
is a reasonable value?

2. If the power consumed is less then 80 watts by a computer, why does
a computer have a power supply rated at 300 watts, for example?

3. Now, here is my calculation:

4 computers = 400VA
1 DSL router = 50VA (I am guessing. Can't be more than that)
1 Network switch = 50VA (I am guessing)

I don't need to protect the monitor.

I saw a 550VA UPS for \$69. Looks like a good deal. Are my calculation
right to make this deal work?

Peter

## Re: How much power does a computer actually draw?

Peter wrote:

I think you're going to get a lot of replies.

My contribution:

1. Your guesses for the router and switch are way high - perhaps
by a factor of 5-10.

2. Power (watts) and volt-amperes are not interchangeable.

3. I generally GUESS typical power required by a desktop alone
to be 100 watts.  I'm weasel-wording that ... the real number
depends on the hardware and what the machine is doing.

4. If you're not going to support the monitor, you're counting
on an automatic orderly shutdown run by something.  The UPS?
How does that work with four machines at once?

5. The right UPS must be capable of enough VA to carry the load
AND enough energy capacity (as in battery size) to continue
carrying that load for however long you want to run before
shutdown.  That's a tricky issue.  There's been a trend in
consumer-grade UPS design to raise the VA rating but stay with
the same size (ampere-hours) battery.  Bigger VA sells well,
but doing it this way means less run time.

## Re: How much power does a computer actually draw?

<snip>

A while back I took some actual measurements on a P-III
and was surprised to see it only took about 50 watts while idling...
but at full cpu usage took a bit over 100 watts...with occasional peaks even
higher.

That was for the machine itself and did not include the monitor etc.

## Re: How much power does a computer actually draw?

Peter wrote:

The UPS web sites may have attempted to query you on the computer
types. For example, power consumption has changed radically as

Some kinds of P3 processor -  35W
Highest P4 processor       - 130W
Low end video card         -  20W
High end video card 8800   - 145W  (time two, if an SLI gaming machine perhaps)

That can mean quite a range of values for a PC. Assuming there
is such a thing as a "typical PC" is a mistake, especially
when you have four of them. Four "high" ones will draw a lot more
than four "low" ones.

A "Kill-O-Watt" meter can make a gross measurement for you. A cheap
one is \$23. I would run Prime95 or a game on each computer, one at
a time, with the Kill-O-Watt meter measuring the power. That will
give you some numbers to base a result on.

But the reality is, if you do things the way you are currently,
I'll see you buying the cheapest UPS you can. If you look at the
UPS manufacturer sites, they have a "lowest line" of products,
that compete on price. This leads to poor quality products.
In a purchase of quantity 100 of these low quality units,
roughly 10% did not function properly (did not protect on
a power outage, did not flip back to line power once an outage
was over etc.). Buying the bottom of the barrel is a mistake
and a waste of money.

When I bought one, I bought the biggest one I could find at a
local store. By buying locally, there was no large shipping
charge (just whatever price they wanted, so I know in advance
what I'm paying). And you can inspect for obvious damage to the
box, before buying it. UPS devices shipped by UPS (brown) delivery,
can arrive crushed due to the weight of the batteries.

For example, here is a UPS for \$24! How long would this last ?

DirectUPS DP400 400VA 200 Watts 2 Outlets UPS
http://www.newegg.com/Product/Product.asp?Item=N82E16842117003

And what are you using for computer monitors ? My old Sony
Trinitron CRT draws 200W. My LCD replacement is much more
reasonable.

Paul

## Re: How much power does a computer actually draw?

wrote:

It might be extremely important to mention the details of
these systems, for example any significant consumers of
power in them since website calculators can be wildly
inaccurate.

What kind of monitor?  It makes a difference.

Does it really "need" to be a single UPS?  I ask because
often the consumer grade around 500VA types are
significantly less expensive per VA, sometimes under \$30 per
on sale or after rebates.  I picked up my last for \$20 but
that sale is over so it depends on how much looking you want

Are you sure you are interpreting this right?  I ask because
it can't be right, you will not have as low as 77W per
system with a monitor, except maybe the most miserly of
energy conservative designs (Via Epia, etc) only a single
drive, and a medium to smaller LCD monitor.  The average
budget OEM system sold today can exceed 77W without a
monitor.  The low end systems' average "might" be as low as
77W if it's mostly sitting idle, but that's not how you'd
want to budget your power, as it certainly can't be sitting
idle in the middle of shutdown sequences.

Why didn't you just mention the system details?
50W is also, probably too low. A conservatively set up
PIII/Celeron system could be around that figure at idle, but
again you'd be looking at the lower power state, not the max
(per system) you'd want to budget for.

There is typically 25%, sometimes more loss in the power
supply.  80% is a current goal, and one many systems don't
meet in best light, let alone when close enough to idle to
draw so little power as estimated above.

You need to find some other website, one modern system with
one monitor should have roughly 500VA consumer UPS.  With a
business/commercial UPS, they are a bit more conservative,
sometimes, and you might be able to budget down closer to
300VA if the momentary peak current is sufficient (else the
UPS just turns off the moment it tries to switch over to
running the systems).

1500-2000VA, depending on exactly what it is.

Because the PSU rating is it's peak output, or if
"sustained", it still means the MAX it can sustain.  You
always want a PSU capable of a little more than you'd
actually use, this applies to any power supply for anything.

However, many PSU labeled as 300W can't even supply 300W,
it's just a daydream put into words stamped onto a label.

On the other hand, PSU also spec higher wattage only because
of the available current per rail, a PSU will try to appeal
to a wider audience by supporting systems with more power
requirement on one rail than another, instead of having to
simultaneously make several different PSU with a similar
total wattage, requiring separate designs, assembly lines,
parts, testing stations, etc.  You are more likely to find a
PSU with a rated wattage closer to the system's peak
hypothetical requirement (perhaps allowing for a slight
upgrade in some cases) with an OEM that can better match
each PSU they spec to a particular system configuration.
Plus, since a PSU manufacturer isn't trying to sell an OEM
unit in retail channels (unless divered to surplus channels,
more conservative about rating it, not trying to be
(marketing/specs) competitive with other retail psu with
agressive, sometimes even fradulent wattage or current
ratings.

But you need it powered still?  So you can't ignore it?

Absolutely not, but you have failed to provide even basic
info about the systems, only data from websites that can't
possibly be right.

Two systems may be reasonably paired with a 600-750VA unit,
sometimes a little smaller if they are old enough systems or
pretty spartan, with only one or two LCD monitors.  Then
with UPS as with any PSU, there's the accuracy of the
manufacturer's ratings and how they can be applied.

If you can get by with a modest 750VA unit for 4 systems
it'll be because it was a commercial unit, essentially
costing as much as a higher VA rated but less robustly
designed unit for consumer or SOHO.  Generally as mentioned
above, you're looking at above 1000VA units just to keep the
UPS turned on, let alone for how long...

You didn't even mention your needed runtime, another
important factor.  With an industrial UPS you might find
larger batteries for some application to extend runtime
compared to the other VA rating, but in general the lower
end products (even from a respected manufacturer) will have
a correspondingly lower runtime at any given load from a
lower VA rated unit.  So suppose you want 30 minutes, a
2000VA unit may not be able to do it, but in all the details
we don't have, lies exactly what you might need for a goal
like that.

## Re: How much power does a computer actually draw?

The power consumption depends ENTIRELY on the components in the computer..
Current dual core CPU based computers with some drives in them usually have
around a 500 Watt power supply unit, and during use on graphics intensive
material draw close to 400 to 500 Watts.  I would reconsider your
calculations.

--
--------
DaveW

## Re: How much power does a computer actually draw?

Buy an ammeter (or a DVM with an amperometer function).  To prevent
having to cut the cord (because the clamp has to go around just one wire
and not both which would cancel each other out), they should include an
inline plug (i.e., you plug the cord into their adapter that then plugs
into the outlet).  This lets you clamp the ammeter around their hollow
adapter so only one wire runs through the clamp.  You then measure
amperage for that device.

You can make lots of guesses or read lots of articles on what might be
typical but what is typical for them isn't what is typical for you.
Also, the amp rating of many devices could be the peak amps (when you
turn the device on or during some particular operation) or it could be
typical amps during normal use.  Also, you have to remember that when
tallying up all the *max* amps for all devices that your power supply is
not 100% efficient.  Most run around 70% efficiency (and why power
supplies also generate heat).  That means if all the parts that are
connected to the power supply consume 10 amps, and if the power supply
is 70% efficient at that particular current draw (they aren't flatline
for efficiency) that the power cord would see 14.3 amps (10 amps / 0.7)
going into the power supply.

http://us.fluke.com/usen/products/category.htm?category=CLMP (FlukeProducts)

The ones that clamp over a wire are usually for measuring high amps.
You could get an extension cord, strip away the insulation in the
middle, cut one wire, and then insert inline a DVM *with* an amps
function (it must be an *A/C* amps function, not the D/C amps function).
Usually DVMs don't measure more than 2 or 10 amps when inline the
circuit.

Of course, you could simply figure on each box drawing its maximum power
and that is determined by whatever power supply is already in the box.
The problem is getting the *true* power rating of the PSU and not what
they may advertise.  The cheapies always overrate their power handling.
If, say, you have a 400W power supply that is 70% efficient then 400W is
all that you should be drawing max (and usually about 75% of that for
many PSUs because of their overratings) then you would be sucking 533W
max into that PSU.  If you had 3 similar boxes then, at maximum, they
would suck in 1600W.  They probably don't draw much more than half that
much but they could if you maxed out your PSU's capacity with lots of
attached devices.  Basically you would figure wattage based on the max
that your PSUs could supply (and remember that they are rated for the
internal consumption by devices so you have to account for their
[in]efficiency).  Then start adding the monitor's max wattage since
obviously an end-user PC that is running but with a dead monitor is of
no use to the end-user.  Stuff like inkjet printers, speakers, scanners,
etc. shouldn't be connected to the UPS since it is highly unlikely
during an outage that you can't do without them.  However, you might
want your router and cable modem on the UPS if you want to surf during
the outage (which can come in handy to lookup your power company's
support number to call in the outage).

If you wants lots of run-time after a power outage, you'll end up
getting pretty big UPS.  I my one home PC with CRT monitor, it takes a
2.2KVA UPS to keep my PC running for about 60 minutes (when the
batteries were new, now about 40 minutes with 3+ year-old batteries).
Maybe all you want is one with enough capacity to keep your system up
long enough to let you close all apps and then gracefully shutdown (or
do so when you're not there).  Rather than shutdown my PC, I have the
UPS software put it into hibernate mode, if the option is available.
Then while you are gone the shutdown doesn't run into apps that won't
exit so the OS pends for them to stop which they don't so the OS gets
slammed off anyway when the power runs out.  Hibernate doesn't ask the
apps to close and wait for them to comply.

Also, you aren't protecting ANY of your hardware with a UPS.  A UPS is
for data protection, not hardware protection.  If you don't have your
monitor plugged into the UPS, what the hell is the purpose of having a
UPS?  Well, that is, you obviously won't be getting an UPS sized to keep
your system up so it is usable to you because you won't have a monitor
to see anything during the outage.  If you are getting a UPS just big
enough to make sure the PCs get shutdown gracefully, yeah, then you
don't need to include the monitor on the UPS.

There is the option of buying a UPS, hooking everything up to it, using
its software to check the percentage of load on that UPS and its guess
at run-time, and if it doesn't have enough reserve for your taste then
return it and get a bigger one.  Of course, you are abusing the retailer
because of your inability, fear, or laziness to actually measure the
current draw on the power cord (with everything running and still giving
yourself some leeway).

## Re: How much power does a computer actually draw?

Thank you all for your help. This has been very information. I have
decided to buy one UPS for each computer, rated at 350VA.

Peter

## Re: How much power does a computer actually draw?

Hey...that's not a bad idea really...
because your entire system will not go down...even if a UPS fails.

I've got a total of three UPS's here...

## Re: How much power does a computer actually draw?

Peter wrote:

Buy one sample of the product you have in mind, then test
it on each computer individually. Pull the plug (or switch
it off via a power strip) and test how well it behaves.
Make sure it is charged fully first, before you do
any testing. And after each power fail simulation,
give it time to recover again as well.

If the design is a dud, better to find out if you only
own one.

Another thing to consider, is the operating principle.
A cheap UPS, connects the input AC to the output AC
under normal circumstances. (I have one and it runs
cool to the touch, because all it is doing is trickle
charging the battery right now.) If the input AC is lost,
the UPS waits to make a decision. It is the "holdup time"
of the power supply inside the computer, that provides
the cushion. That is why the power supply in the computer
has that 470uF capacitor in it. The capacitor holds
enough charge to keep the computer running for 15 to 20
milliseconds. The UPS then switches to the battery
powered source, and powers the computer (hopefully,
before the 470uF capacitor is drained). Such a UPS is
called an SPS or Standby Power Supply. It isn't really
a UPS, because the output is actually interrupted for the
switching time. It needs enough time to decide whether
the power has really failed, and it cannot do that in
zero time (if, say, the waveform is merely distorted
for a half cycle).

If a computer power supply does not have sufficient
hold up time, a UPS (SPS) won't fix it. The computer
will do a reset, before the UPS has finished switching
over.

See the "Standby (Offline)" description here, to see
what you are getting four copies of. Then read the
description of the "Line Interactive", to see what
you could be getting instead. The "Dual conversion online"
used to be the expensive one, as it did AC-DC-AC
conversion, which means all "sins of the AC network"
are filtered out. They used to be \$1000-\$1500.

http://en.wikipedia.org/wiki/Uninterruptible_power_supply

I'm not really sure an SPS is what you want. That is
going to be a "feel good" purchase.

I own an SPS, to correct for the "line slap" and
sub-second outages I get on windy days, in my residential
setting. And the occasional one second outage from
some kind of upstream switching. And I have enough
run time, to do an orderly shutdown, and that is about
all.

Paul

## Re: How much power does a computer actually draw?

On Fri, 16 Feb 2007 23:48:18 -0600, "Vanguard"

The other problem is power factor, depending on where the
"weakest link" is in the UPS, it's max VA rating may or may
not be able to sustain the higher momentary current from an
average power estimation or RMS figure.  Then there's the
matter of whether the UPS outputs sine or square waves.

There's also the matter of inrush current with an offline
UPS, during the moment the UPS kicks in and power was
interrupted, the PC PSU is draining it's caps, there will be
a higher (than later) initial surge to recharge them.

Essentially, a whole lot of fudging and buying more than a
minimal value UPS, can not only avoid doing tons of
calculations, but also helping to offset any overly
optimistic ratings by the UPS manufacturer, and hopefully
resulting in enough margin that the UPS isn't being pushed
to the max the design can provide which with most power
circuits can tend to reduce it's lifespan, even unduly drain
the batteries fast enough that if the location has frequent
outtages, it can be just as expensive over several years to
replace batteries more often to retain acceptible runtime.

On the other hand, some UPS batteries are more common and
cheaper to source from 3rd parties like the typical 12V, 7Ah
and 12Ah types may have substantially higher Ah/\$ at roughly
\$15/\$25 respectively.

## Re: How much power does a computer actually draw?

Peter wrote:

A UPS will solve this if the problem is caused by momentary dip in line
voltage.  It will not cure a software bug.  Are you certain it's a power
problem?

## Re: How much power does a computer actually draw?

Peter wrote:

... snip ...

At least in the US and Canada you don't need any fancy metering
instrumentation.  Just turn off everything in the house, and check
that the outside wattmeter is quiescent (no wheel rotation).  Turn
on a calibrated (more or less) power drain, such as a 100 Watt
bulb, and time one rotation of the wheel.  Now you can turn on the
computer, turn off the bulb, repeat the measurement, and easily
calculate the drain of the computer.  With a few simple equations
you can handle any draw in the presence of any constant background
drain.

Rotation speed is proportional to power draw, and inversely
proportional to the time per rotation.  That's all you need.

--
<http://www.cs.auckland.ac.nz/~pgut001/pubs/vista_cost.txt
<http://www.securityfocus.com/columnists/423

"A man who is right every time is not likely to do very much."
-- Francis Crick, co-discover of DNA
"There is nothing more amazing than stupidity in action."
-- Thomas Matthews

## Re: How much power does a computer actually draw?

Peter wrote:

... snip ...

How are you going to bring down the machines without a monitor, in
case of a long term power outage.  The neighbors kid just knocked
over a pole, or an ice storm hit.

--
<http://www.cs.auckland.ac.nz/~pgut001/pubs/vista_cost.txt
<http://www.securityfocus.com/columnists/423

"A man who is right every time is not likely to do very much."
-- Francis Crick, co-discover of DNA
"There is nothing more amazing than stupidity in action."
-- Thomas Matthews

## Re: How much power does a computer actually draw?

Thank you all for your suggestions.

The problem that I have is a wierd one. Occassionally, one or two
computers just reset themselves. The circuit breaker itself does not
trip. I was told that the surge protector I am using helps the circuit
breaker not to trip.

There is definitely some problem in the circuit. I want to use the UPS
to avoid those temporary surges (or whatever that thingy is).

Peter

## Re: How much power does a computer actually draw?

Hmm, that's only paraphrasing the very much older expression: