Define the machine
Record the Windows build, CPU, GPU, driver, power state, and background conditions.
HOW WE TEST
Confirm the setting. Then measure the workload.
A Windows value can confirm that a control changed. Performance still needs a repeatable comparison on a defined machine.
THE TESTING RULE
Change fewer variables, repeat the same path, and keep the full setup beside the result.
Use the same method to compare one MOLDS control on your own hardware.
Fix the workload
Use the same scene, route, duration, warm-up, and run count for every comparison.
Test one change at a time
Keep the selected MOLDS control attached to the run so the difference remains explainable.
Compare distributions
Look at frame times and the 1% and 0.1% lows, not only the average. CapFrameX, one of the Useful Programs, records them.
WHAT A RESULT NEEDS
A number is useful when its conditions travel with it.
A published comparison names the test setup and the controls that were on, so anyone can repeat the run.
SystemCPU, GPU, memory, driver, and Windows buildWorkloadScene, path, warm-up, duration, and run countMetricsAverage, low percentiles, and frame-time distributionChange setThe exact controls active for that comparison