A review is only worth reading if you know how the numbers were made. This page is the method, kept in one place so we do not have to repeat it in every article — and so you can tell us when it is wrong.
The rules
- Every machine is tested at 22 °C ± 1, on the same desk, with the same ambient noise floor.
- Every benchmark runs three times. We publish the median, never the best.
- Battery tests start at 100% and end at shutdown, at 150 nits, with Wi-Fi on.
- Nothing is reviewed on a pre-release driver unless the manufacturer ships it to buyers.
What we measure
- Sustained CPU load, 30 minutes
- Single-core responsiveness
- Display brightness and colour
- Fan noise at the ear
- Charge time, 0 to 80%
What we do not
- Synthetic scores with no unit
- One-off peak numbers
- “Feels fast” impressions in a scored table
- Anything we cannot repeat next month
Why the median, not the best run?
Because a laptop that hits a high score once and then throttles is not a fast laptop — it is a laptop with a good first minute. The median is the number you will actually live with.
The bench, in numbers
| Test | Duration | Runs | What it tells us |
|---|---|---|---|
| Sustained load | 30 min | 3 | Cooling headroom |
| Responsiveness | 8 min | 3 | Everyday feel |
| Battery, mixed use | To shutdown | 2 | Real working day |
| Display | 15 min | 1 | Brightness, gamut, uniformity |
| Noise | 30 min | 3 | dB at 30 cm, under load |
Why we do not score things out of 100
A single number hides the trade-off that made the machine interesting. We give a verdict and the measurements behind it, and we let the trade-off stay visible.
Read the reviews
Found a mistake in the method? Tell us — corrections are published on this page, dated.
Would like to see the noise measurement at idle too, not only under load.
Publishing the median instead of the best run should be standard everywhere. Thank you.