The short answer
A PSU's wattage rating describes its output capacity, not a constant electricity draw. Price the computer's measured wall-input kWh or a stated average for each workload and power state. Include monitors separately unless a group reading already captures them, and do not add PSU conversion losses to energy already measured at the wall.
Use the electricity cost calculatorOn this page
A 750 W PSU does not establish 750 W all day
The power supply is sized to deliver power to the computer’s components. Its rated output is a capacity, not a report of the average AC input during your workday. For example, Corsair’s HX1200i specification labels its output and total capacity explicitly. That 1,200 W product declaration is not evidence that an ordinary desktop takes 1,200 W continuously.
Seasonic’s PSU-sizing guidance also distinguishes maximum-load assumptions from real workloads. A PSU-selection calculator estimates capacity needed for hardware; an electricity-cost calculator answers a different question about energy over time. Avoid entering the recommended PSU size as though it were measured average power.
Define the setup before measuring
Decide whether your boundary is the tower alone or the entire desk. A monitor, speaker system, powered USB dock and network device may have their own sockets. USB equipment powered through the computer is already inside a tower wall-input measurement when connected.
| Available number | What it can support |
|---|---|
| Accumulated kWh at the tower’s socket | Energy of the tower and its included powered accessories during that period |
| A measured whole-desk power strip | Combined use of the equipment actually connected |
| CPU or GPU software power telemetry | That reported component or subsystem, not automatically the whole AC setup |
| PSU output capacity or a sizing recommendation | Hardware capacity planning, not a measured daily cost |
Keep hardware, display settings and the workload with the reading. Gaming, office work, rendering and idle time need not have the same average. A benchmark under full load is not automatically representative of an eight-hour workday.
A mixed-workload example
The following powers are illustrative assumptions, not typical values assigned to desktop PCs or a test of any build. They describe wall-input power for the tower only:
| State | Assumed average | Hours/day | Energy/day |
|---|---|---|---|
| Demanding work or gaming | 250 W | 3 | 0.750 kWh |
| Lighter work | 70 W | 5 | 0.350 kWh |
| Sleep state | 3 W | 16 | 0.048 kWh |
| Total | Different averages by state | 24 | 1.148 kWh |
Repeated on all 365 days, this gives 419.02 kWh/year, costing $83.80 at an illustrative $0.20/kWh. If that routine occurs on fewer days, model the other days separately rather than automatically using 365 workdays.
Edit the 250 W demanding-use portion. That link prices just one row. Add the compatible state totals once in the household planner or use the standby tool for low-power hours.
Do not apply the efficiency adjustment twice
Electricity supplied at the wall includes PSU conversion losses. A plug monitor’s 0.75 kWh reading is already an AC-input energy figure; adding another assumed 10% for the PSU would double count a loss already captured.
For an explicitly different DC-output scenario, suppose the components need 300 W and efficiency at that operating point is assumed to be 90%. The estimated AC input would be 300 ÷ 0.90 = 333.33 W. This is illustrative conversion arithmetic, not the efficiency or consumption of your PSU. An efficiency badge does not provide one exact efficiency at every load or input voltage.
For a household estimate, accumulated wall kWh over a representative routine generally avoids having to estimate every internal component and conversion loss. Use an appropriate meter according to its instructions; there is no need to open the PC or power supply.
Sleep, shutdown and display-off are different
Microsoft documents different Windows sleep and hibernation behavior, with availability depending on hardware. A dark monitor does not by itself establish that the tower is asleep, and a configured timeout may not describe the state reached during downloads or another active task.
Compare the states your computer actually reaches. In the journal, retain complete-day observations before and after a settings change with the same workload boundary. Describe any difference as observed under those conditions; it does not prove a universal saving for other computers.
Bring the result back to the bill
Use your own price per kWh in USD, GBP or EUR on a consistent tax and billing basis. The tariff guide explains fixed fees and unit prices. Where a desktop and monitor were metered together, add that group once, rather than adding the monitor again.
The bill-period appliance check helps compare the matching period’s device estimates with billed kWh. A remaining gap can reflect other loads, routine differences or incomplete boundaries; it does not automatically prove a PC or meter problem.
Price separate operating states
Choose “Separate active, idle and standby modes” in the electricity calculator. Record average input watts, daily hours and a source or assumption for each state. The hours must total no more than 24; unassigned time remains excluded. A complete-day measurement is an alternative and must not be added to overlapping mode estimates. Follow a reproducible mode example before applying your own tariff and use days.