Test protocol
Method
This is how the tests behind every result are run. Each claim on the channel carries one of the labels below.
Labels
- VerifiedMeasured or checked on my own unit, under the conditions shown.
- InferenceMy reading of the data. I say so when that's what it is.
- UnverifiedMaker and press figures. They stay here until I've measured them.
- VerdictMy weighted buyer score for a video.
- PendingPlanned, or still running.
Protocol
- Capture
- CapFrameX records the frame data. HWiNFO64 logs the sensors with its own FPS counter switched off, so it doesn't clash with PresentMon. RTSS draws the overlay, and OBS records on a separate PC so encoding never runs on the device being tested.
- Power
- Power limits (SPL, SPPT and FPPT) are matched in each device's own software, then confirmed in HWiNFO64. Wall power is metered for every setup.
- Conditions
- Driver versions are locked and shown on camera. CPU boost is off for gaming runs, after an A/B check, except in CS2. Downloads and background tasks are paused while a run is measured, and FurMark always runs last.
- Runs
- A summary number only appears after three runs. Anything with fewer is marked provisional, with its run count. When two machines are compared in games, both run at the same time.
- Upscaling
- Internal render resolution is disclosed. If a game doesn't expose it, I say so instead of guessing.
- Noise
- Fan noise is measured in dBA with a calibrated microphone.
- Money
- I buy everything I review, and the opinions are my own. Every video description says so.
The capture tools are listed on the links page.