How MacBookBatteryMonitor measures power: 99.63% accurate

Accuracy across every run

We scored MacBookBatteryMonitor and four traditional ways of reading Mac power against the Mac's own system power sensor, across 7 runs on an M1 Max MacBook Pro and an M2 Mac mini. After its 15-minute calibration, MacBookBatteryMonitor counted the energy each Mac used with 99.63% accuracy. The best traditional method, every chip power sensor added up, reached 62.53%.

powermetrics reports CPU + GPU + Neural Engine. Stats' power sensors add DRAM and PCIe to that. CPU power × 2 is the fixed factor WattMate publishes for turning processor energy into battery draw. Every chip power sensor added up is everything the chip reports about its own power.

The MacBook Pro working session, minute by minute

This is one real working session on a 16-inch MacBook Pro with M1 Max (MacBookPro18,4, macOS 15.7.7): 2,972 seconds of ordinary work. One-minute averages of the Mac's power swung from 12.34 W to 43.92 W, and it drew 23.70 W on average.

After the first 15 minutes, the Mac drew 13.39 Wh. MacBookBatteryMonitor counted 13.33 Wh, 99.52% of it. Apple's powermetrics saw 3.49 Wh, 26.06%. Across the whole session powermetrics showed 6.47 W on average while the Mac drew 23.70 W.

Look at where the gap sits. At minute 3 the Mac drew 43.92 W and Stats' sensors read 20.43 W. At minute 13 the Mac drew 12.34 W and powermetrics read 1.08 W. The share each method sees changes minute to minute, so no correction factor applied afterwards can fix it.

Every run, every method

Each row is one run, scored after MacBookBatteryMonitor's first 15 minutes. MacBookBatteryMonitor started each run from the same generic values and adapted to each Mac live, so every row is a fresh calibration. Alongside energy accuracy, the table shows how closely MacBookBatteryMonitor tracked each one-minute average and each one-second reading.

How MacBookBatteryMonitor measures power

The total comes from the Mac's own power sensor

Every Apple silicon Mac has a system power sensor in its System Management Controller (SMC key PSTR) that reports what the whole machine draws. MacBookBatteryMonitor reads it every second and pairs each reading with the inputs from that same second. That measured figure is the total you see.

23 inputs, 16 components, one measured total

Underneath the total, a model takes 23 inputs across the chip's power sensors, storage, network, fans, displays and USB. It turns them into 16 components: CPU, GPU, Neural Engine, DRAM, memory controller, fabric, media, ISP, display controller, PCIe, SSD, network, fans, display, USB and logic board.

Every second, MacBookBatteryMonitor reconciles those 16 components so they sum exactly to the measured total. The Calibrated view has no unexplained "other" bucket.

Adaptive calibration tunes itself to your exact Mac in 15 minutes

Every Mac turns the same inputs into a different number of watts. MacBookBatteryMonitor's adaptive algorithm tunes itself to each Mac's power profile live, starting from generic values, and dials it in over the first 15 minutes of use. From then on it counts the energy your Mac uses with 99.63% accuracy.

The lab runs were scored exactly the way the app runs on your Mac. Each second was predicted before the model learned from it, so no score on this page comes from data the model had already seen.

Raw Sensors view

Switch to Raw Sensors and MacBookBatteryMonitor lists every sensor exactly as macOS reports it, plus a Not Covered line: the power that the chip's own sensors do not account for. It is the gap every chip-only tool leaves out, shown live.

Per-app watts

MacBookBatteryMonitor splits the measured total between your apps by how much work each one does, so every app's share adds up to what your Mac actually drew.

• CPU watts per app come from the kernel's per-process energy accounting, scaled to the measured CPU power for that second.
• GPU watts per app come from each process's GPU usage, scaled to the measured GPU power.
• Helper processes roll up into the app responsible for them, so every Chrome helper counts as Chrome.
• Short-lived processes, such as build and test workers that start and exit between samples, are counted.
• Docker containers, cloudflared and launchd services are tied to the process hosting them.
• The list refreshes every 2 seconds by default, and you can choose any interval from 1 s to 60 s.
• Per-app charts cover 1 hour, 8 hours, 24 hours and 7 days, and the app keeps 30 days of every-process data on your Mac at full resolution, including processes under 10 mW, on battery and while plugged in.

Why traditional methods miss

Chip sensors stop at the chip

powermetrics, and every app built on the same chip power sensors, reads power inside the system on a chip. Your Mac draws a large share of its power outside it: the display, SSD, Wi-Fi, fans, USB, power delivery and the logic board. None of that appears on a chip sensor. Every chip sensor added up reached 62.53% across all runs, and on the MacBook Pro it counted 6.89 Wh of the 13.39 Wh the Mac drew.

Doubling CPU power assumes a ratio the hardware doesn't hold

WattMate's How we count page counts every joule of processor work as 2 joules from the battery, on every Mac, and calls the factor "a rounded compromise". The real ratio moves with the workload. On the MacBook Pro session the Mac drew 2.72 times its CPU power at minute 1 and 11.98 times at minute 13. On the M2 Mac mini real-tasks run, CPU × 2 counted 3.30 Wh against a measured 3.18 Wh, too high. On the MacBook Pro it counted 4.43 Wh against 13.39 Wh, far too low. Pooled across every run, CPU × 2 reached 61.31%.

Stats reads the right total and an incomplete breakdown

Stats is open source, so anyone can check exactly what it reads. Its System Total sensor reads PSTR, the same system power sensor MacBookBatteryMonitor uses for its total. Its component power sensors come from five of the chip's energy channels: CPU, GPU, ANE, RAM and PCI (Modules/Sensors/readers.swift). Those five channels saw 51.22% of the energy across all runs. On the MacBook Pro they counted 4.72 Wh of 13.39 Wh, and Stats lists them beside the total without reconciling the gap between them.

Stats' per-app "power" column runs top -o power (Modules/Battery/readers.swift). That column is macOS's Energy Impact score, a unitless number. It is not watts.

Activity Monitor shows a score

Activity Monitor's Energy tab ranks apps by Energy Impact, the same unitless score. Its 12 hr Power column is an average of that score. Neither one is measured in watts.

How we tested

The lab has recorded 33 runs and 39,248 seconds (10.9 hours) on two Macs. The scores on this page come from 7 runs: 16,849 seconds recorded, of which 10,549 were scored after MacBookBatteryMonitor's first 15 minutes.

Reference

The reference is the Mac's system power sensor (PSTR), verified by measuring power draw from the battery and from the device. Every second, each method's reading is paired with the PSTR reading for that same second.

Workloads

The 7 scored runs cover four workloads: the calibration protocol (2 runs on the M2 Mac mini), a CPU × GPU validation grid (2 runs), real tasks (2 runs) and a full working session on the M1 Max MacBook Pro (1 run).

The calibration protocol

The protocol loads each component on its own at stepped levels while the system power sensor is recorded every second. Alongside it, the lab records the chip's energy channels, fans, temperatures, battery telemetry, CPU load, and disk and network throughput. Combined CPU + GPU steps and a CPU × GPU grid are held out of the fit, so they test the model on loads it never trained on.

Scoring

MacBookBatteryMonitor started every run from generic values, with no per-model tuning, and predicted each second before it learned from it. The headline figure is energy accuracy: 100% minus the difference between the watt-hours a method counted and the watt-hours the Mac drew, after the first 15 minutes of each run. Every method is scored the same way over the same seconds. Pooled across all 7 runs, MacBookBatteryMonitor reached 99.63%. It also tracked one-minute averages with 99.19% accuracy and individual one-second readings with 98.57% accuracy, scored as 100% minus the mean absolute percentage difference.

See your Mac's real power draw

MacBookBatteryMonitor shows your whole Mac's power, what it is going to and the watts every app is using, live in the menu bar. It also slows down or offloads a power-hungry app without quitting it.

Compare the methods app by app

• MacBookBatteryMonitor vs Stats
• MacBookBatteryMonitor vs WattMate
• MacBookBatteryMonitor vs BatteryBud
• MacBookBatteryMonitor vs PowerFlow
• Best Mac power monitor apps