How to Measure the Power Draw of Your Equipment

The number on a power supply label is a ceiling. The number in a hardware monitoring app is a partial view. The number you need for UPS sizing is the real AC power flowing from the outlet into the equipment, and there are only a few ways to get it. This page covers each tool, what it actually measures, where it lies to you, and a protocol that turns readings into a UPS decision.

What you are trying to capture

A UPS supplies AC power at its outlets, so the useful quantity is AC watts at the plug. Three numbers matter:

• Idle: the system on, doing nothing. This drives long-runtime planning and energy cost.
• Sustained load: the highest steady draw during your heaviest realistic task (a game, a render, a NAS scrub). This drives capacity sizing.
• Transients: spikes lasting microseconds to milliseconds, mostly from graphics cards and CPUs changing state. These can trip a UPS overload even when the sustained reading looks safe.

The first two you can measure. The third you mostly cannot with consumer tools, which is why headroom exists.

The tools, ranked for UPS sizing

Plug-in watt meters

These "Kill A Watt style" meters plug into the wall and accept the device's plug. Inside, they sample voltage and current many times per cycle and compute true RMS values, real power and power factor. For sizing, they are the default tool.

Know their limits. Most update the display roughly once per second, and the value shown is an average over that window. Accuracy is typically within a few percent for loads from tens to hundreds of watts, but the reading at a few watts can be unreliable. The current limit is usually 15 A, which is 1,800 W at 120 V. Some models have a "max watts" memory; if yours does not, watch the display during the heaviest part of the workload and note the highest number.

Energy-monitoring smart plugs

A smart plug with power monitoring logs watts and kWh to an app or to a home automation system. That makes it the best way to learn the real average over days, which matters for energy cost and for runtime at typical load. Update intervals of several seconds or more mean peaks are smoothed even more than on a plug-in meter. Some smart plugs are rated below 15 A, so check before putting a large PC on one.

UPS front panel and software readings

If the equipment is already on a UPS, its display or software shows the load. Useful, with caveats. Some units show a bar graph in coarse steps. Some report load as a percentage of the watt rating, others as the higher of watt and VA percentages, and the method varies by brand; UPS load percentage explained covers the differences. The sensing circuits exist mainly for overload protection, and on many consumer units a reading of a few percent can be off by a wide margin. Treat it as a cross-check rather than the primary measurement.

Software such as the maker's utility, Network UPS Tools or Home Assistant can log the reported load over time, which is handy for spotting peaks you missed.

Clamp meters, and why a whole cord reads zero

A clamp meter senses the magnetic field around a conductor. A power cord contains the hot and the neutral, carrying equal current in opposite directions at every instant. Their fields cancel, so a clamp around the whole cord reads zero or close to it, regardless of load.

To use a clamp meter on a plug-in device, you need a line splitter: a short adapter that separates the hot conductor into its own loop. Many splitters have a second loop that passes the conductor around ten times, multiplying the reading by 10 for better resolution at low currents. Divide that reading by 10.

Even then, most clamp meters measure amps only. Amps multiplied by voltage gives VA, not watts. For a modern PC with power factor near 0.98 the difference is small; for small adapters with a power factor of 0.6 the VA figure can overstate real power by more than half. Hardwired circuits and panels are work for an electrician.

PSU and component software: DC side vs AC wall

Monitoring tools report CPU package power, GPU board power and sometimes, with a digital power supply, PSU output. All of these are DC-side values. The wall sees more, because the power supply loses some energy as heat, and because software readings rarely cover the motherboard, memory, drives, fans and USB devices.

Example: software reports GPU 280 W and CPU 120 W, so 400 W. Allow roughly 50 W for everything else, giving 450 W of DC load. A power supply around 90% efficient at that load pulls 450 / 0.90 = 500 W from the wall. The software total of 400 W understated the UPS load by 20%. Use software to understand where power goes, and a wall meter to size.

A measurement protocol

Run this once per machine or per desk. It takes 20 to 40 minutes.

1. Connect the meter at the wall, with the device or a power strip of devices plugged into it. Everything you plan to put on battery should go through the meter.
2. Let the system settle for a few minutes after boot, then record idle watts, VA and power factor.
3. Run your heaviest realistic workload for at least 10 minutes: the most demanding game at your usual settings, a full render, a backup plus scrub. Add a stress test only if you actually run that kind of load.
4. Record the highest sustained reading, ignoring a single odd blip but not a value that repeats.
5. Check monitors at your real brightness. HDR content and max brightness can raise a monitor's draw noticeably.
6. Note startup draw for drives and NAS units during spin-up, which can briefly exceed steady draw.
7. For averages, leave a kWh-accumulating meter or smart plug in place for 24 hours or a full work week.

Worked example: from reading to UPS load %

A workstation desk goes through one meter. Results:

The candidate is a 1000 VA / 600 W UPS. Watt load: 412 / 600 = 68.7%. VA load: 425 / 1000 = 42.5%. The watt figure is the binding one, as it usually is with modern gear.

68.7% sits inside the 50 to 80% planning band, but this desk has a discrete graphics card whose spikes the meter cannot see. Adding a 20% allowance gives 412 x 1.2 = 494 W, or 82.4% of 600 W: too tight. A 1500 VA / 900 W unit puts the measured peak at 45.8% and the padded figure at 54.9%. That is the better fit.

The idle figure informs runtime. If the user walks away during an outage, shutdown software sees the PC near 118 W, not 412 W, and runtime at that load is several times longer. The runtime calculator will show both.

Turning kWh into average watts and cost

If a meter shows 2.4 kWh after exactly 24 hours, average power is 2.4 x 1,000 / 24 = 100 W. Over a year that is 100 W x 8,760 h = 876 kWh. At roughly $0.17 per kWh (around the US residential average) that costs about $149 a year; at $0.30 per kWh, about $263. The UPS energy cost calculator adds the UPS's own losses, discussed in UPS efficiency and energy cost.

Diagnosing odd readings

• Reading jumps between 0 and a few watts: normal for very small loads like a switch or ONT. Measure them together on a strip.
• Power factor around 0.5 to 0.7 on a computer: an older power supply without active PFC. VA will be much higher than watts, so check the VA limit too; active PFC power supplies explains the difference.
• Voltage reading far from 120 V: the meter also shows line voltage. Readings regularly outside roughly 114 to 126 V are worth investigating; see measuring power quality at home.
• UPS and meter disagree by 10 to 20 W: expected. Trust the meter for sizing.
• Draw higher with the UPS in line than without: the UPS has its own losses. Measure the equipment, not the UPS input, when sizing.

For 230 V regions the method is identical; meters are sold for local plug types, and a 10 A or 16 A limit is common. Watts are watts, so the UPS load percentage works out the same way. Feed your numbers into the UPS sizing calculator or the full sizing method, and compare with typical figures in the device power draw database.