How many watts does a computer actually use?

The short answer: a normal desktop draws about 16 watts sitting idle and somewhere between 50 and 150 watts doing ordinary work. A laptop on a dock draws around 5 watts idle, maybe 35 under load. A 24–27” monitor draws about 16 watts with the screen on.

If you’ve searched this before, those numbers will look wrong. Nearly every page on the subject says a desktop uses 200–500 watts. That figure is real, but it isn’t a measurement of anything — and knowing where it comes from is the whole point of this post.

Where “200–500 watts” comes from

It’s the power supply rating. The number on the sticker inside the case.

A power supply is rated for the maximum it can deliver, not what the machine asks of it. Buying a 650 W supply doesn’t make your PC draw 650 W any more than a 120 mph speedometer makes your car do 120. It’s a ceiling, and the ceiling is deliberately generous — you size a power supply for the worst moment the machine will ever have, then add headroom on top.

So when a page tells you “desktops use 200–500 watts,” it is reporting a component specification and calling it consumption.

What the measurements say

Here’s the useful part. The US EPA publishes the ENERGY STAR certified product database, and to be certified a computer has to be measured on a test bench. That data is public domain, and it covers 1,818 machines.

I pulled it. Median idle draw, by machine type:

Type	Machines measured	Median idle	Busier end
Desktop	297	10.7 W	25.5 W
All-in-one	126	15.9 W	21.7 W
Workstation	70	70.1 W	165.4 W
Laptop	1,180	4.9 W	7.0 W
Thin client	10	4.7 W	7.3 W

“Busier end” is the 90th percentile — nine in ten machines of that type sit below it.

And the part worth sitting with. 161 of those desktops publish both their power supply rating and their measured idle draw. For those same machines:

• median power supply rated: 330 W
• median measured idle: 16.5 W
• median ratio: 21.7×

The label is, on the typical certified desktop, about twenty-one times the machine’s actual idle draw.

⚠️ Idle is not the whole story — and I’d be doing the same thing if I stopped here

If I left it at “desktops use 16 watts,” I’d be making the identical mistake in the opposite direction.

ENERGY STAR only measures idle states. Short idle, long idle, sleep, off. There is no “running a build” or “playing a game” number anywhere in that dataset, because that isn’t what the test is for. A machine doing nothing tells you the floor, not the range.

Realistically:

• Office desktop doing office things — 50 to 150 W. Web, mail, documents, a few tabs.
• Same machine idle — the ~16 W above.
• Gaming PC under load — 300 to 500 W is entirely plausible, and this is the honest part: for that specific case, the number everyone quotes is roughly right.

Which is the actual problem with those pages. 200–500 W is a gaming-PC-under-load figure being served as the answer to “a computer.” It’s not invented, it’s just wildly out of context — and it’s the number people then use to size a UPS, estimate a power bill, or decide whether to leave a machine on overnight.

Monitors, while we’re here

Same pattern, smaller numbers. Folk wisdom says a monitor is about 30 W. Measured on-mode power from the same database:

Screen size	Monitors measured	Median on-mode
20–23”	555	12.5 W
24–27”	613	15.9 W
28–31”	123	26.9 W

A 27” monitor is a 16-watt device. Two of them cost you about as much as one old incandescent bulb.

How to measure your own in about a minute

None of the above beats measuring the machine in front of you, and it’s cheap to do.

A smart plug that reports true watts — a TP-Link Tapo P110/P115 or similar — costs less than being one size wrong on a UPS. Put the machine through it and read the number.

Two things that will skew you if you’re not careful:

• Measure at normal working load, not at the login screen. An idle reading under-sizes anything you plan from it.
• Watch it for a while. A desktop’s draw is spiky. The peak matters for sizing; the average matters for your bill.

Already own a UPS? Most with a display will show the current load in watts. Cycle to the LOAD/WATTS screen and you have the whole rack’s draw for free. Not the VA screen — VA is a bigger number that measures something slightly different, and using it here would over-state your draw.

What it costs you

Once you have a real watt figure, the arithmetic is the easy part:

kWh per year = watts × hours per day × 365 ÷ 1000
cost         = kWh × your rate

An office desktop averaging 60 W, on eight hours a day, is about 175 kWh a year — roughly $30 at 17¢/kWh. The same machine left on 24/7 is about 525 kWh, or $89. A gaming rig averaging 300 W for four hours a day lands near $75.

That gap between “on while you use it” and “on all the time” is usually worth more than any component you’d swap to save power.

(ENERGY STAR does publish an annual energy figure per model — median 41.9 kWh for a desktop, about $7 a year. Don’t use it for this. It assumes a fixed test pattern in which the machine is asleep or off roughly 60% of the time, which is fair as a comparison standard between products and nothing like how most of us actually use a computer.)

Where these numbers come from

Everything measured above is from the ENERGY STAR certified product database, published by the US Environmental Protection Agency and in the public domain — you can download it yourself:

• Computers: data.energystar.gov dataset rxdj-2c88 (1,818 models)
• Displays: dataset qbg3-d468 (1,588 monitors)

Two honest caveats. It’s a certified list, so it skews toward efficient models — the installed base probably runs a little higher than these medians. And as above, it measures idle, not work.

If you want to put this to use, I built a UPS sizing and runtime calculator on top of the same data. It uses these measured figures where they exist, says “ballpark” where they don’t, and shows its working rather than asking you to trust it.