The electric bill lands, it's higher than you'd like, and your eyes go straight to the PC humming in the corner. So you ask the obvious question: how much is that thing actually costing me? You glance at the power supply, see "750W," and figure that's your answer.
It isn't. That 750 doesn't mean your PC draws 750 watts, any more than a 240 km/h speedometer means your car does 240 in traffic. The label is a ceiling — the most it could ever pull — not what it's pulling right now. The real number is hiding, and it's almost never what you'd guess.
Good news: you can find it. This post shows you three ways to measure what your PC really draws — one free, one you might already own, one for the price of lunch — and tells you exactly what each one gets right and what it quietly misses.
The number nobody knows
A PC's draw swings by an order of magnitude depending on what it's doing:
- Laptop, browsing and office work: roughly 10–40 W.
- Desktop, sitting idle: typically 50–100 W at the wall.
- Gaming desktop under load: 300–600 W is normal for a high-end build.
Same machine, ten-fold range. "My PC uses X watts" isn't really a sentence that means anything — it uses X watts right now. And Windows won't tell you: Task Manager charts CPU percent, memory, even GPU queues, but there isn't a single watt anywhere in it.
Three ways to get the number
- A wall meter — the truth, for the price of lunch. A plug-through power meter (the classic Kill A Watt style) or any smart plug with energy monitoring sits between the wall and your PC and reports exactly what flows past, including everything software can't see: motherboard, RAM, drives, fans, RGB, and the power supply's own conversion losses. Measure three moments — idle, normal work, full load — and you know more about your machine than most owners ever do. This is the method to use if you need numbers you'd put in a spreadsheet.
- The UPS you already own. Most consumer UPSes display output watts on their panel or in their companion software. It's coarser, and it lumps together everything plugged into the battery side — but if a UPS is already under your desk, it's a free wall meter.
- Software sensors — free, instant, and half the picture. Modern CPUs report their own "package power" and NVIDIA GPUs report board power; tools like HWiNFO or LibreHardwareMonitor read both. Two caveats matter. On a desktop, CPU + GPU is not the whole machine — the missing motherboard, RAM, drives, fans and PSU losses mean the wall figure is routinely 30–60% higher than what the sensors add up to. On a laptop running on battery, though, there is one genuinely honest software number: the battery's discharge rate is the whole machine, screen included, measured by the hardware that's actually paying for it.
Turning watts into money
One formula is all you need:
kWh per year = watts × hours per day × 365 ÷ 1000
A desktop that idles at 120 W and never sleeps burns about 1,050 kWh a year — roughly $160 at US average rates, and well over €300 at European ones. That's for doing nothing.
But the real value of measuring isn't computing an average — it's finding the waste window: the hours when the machine runs at full-power settings and nobody benefits. A PC drawing 120 W overnight when it could be drawing 40 in a capped power mode — or 2 in sleep — is the gap all the savings live in. (If your machine genuinely can't sleep because it serves Plex or torrents, that gap is still closable — we wrote a separate guide for always-on machines.)
Measurement built in, next to the fix
Knowing the number saves nothing by itself; the savings come from switching states — full power when you need it, capped when you don't. That's why PowerDoze builds the measurement into the same app that acts on it:
- Live watts on the dashboard. On a laptop on battery, that's the real whole-machine figure from the battery's discharge rate. On a desktop or plugged in, it's the CPU + GPU sensor sum — and the app labels which kind you're looking at instead of pretending they're the same thing.
- Per-mode history. PowerDoze logs power samples per power mode, so "how much does Quiet mode actually save vs Full Power" gets answered with your machine's own data, not a brochure number.
- Savings vs your own baseline. The savings stats compare against the most power-hungry mode you've configured — and they're tagged as real measurement only when the data came from the battery, estimate otherwise.
Then the automation — schedule, away detection, fullscreen and per-app rules — does the switching, which is the part no meter can do for you.
Honesty corner: what software numbers are good for
Software estimates are estimates. On a desktop, the sensor sum undercounts the wall by design, GPU readings cover NVIDIA cards (AMD and Intel GPUs don't expose the same interface), and no app — ours included — can see your PSU's efficiency curve. If you need precise numbers for billing or reimbursement, buy the wall meter; it costs less than a month of the waste it will find.
But for saving power, precision isn't what matters — direction is. "This mode cut 35 W off that mode" is true in software even when both absolute numbers run low, and trends over weeks are exactly what tells you whether a change worked. Use the meter to calibrate once; use the software to watch every day.