Performance testing measures frame rate, memory, load times and thermals under realistic and worst-case conditions. Soak testing leaves the game running for hours to surface the slow failures — memory leaks, handle exhaustion, numeric overflow — that a twenty-minute session never reveals.
Useful performance testing produces numbers, not impressions. "The village feels laggy" is not actionable. "Frame time rises from 16ms to 38ms when entering the village from the north gate, with the spike in the draw call submission, reproducible every time" tells a programmer where to look.
Soak tests are deliberately boring and deliberately long. You leave the build running in a loop — idling in a level, or driving a bot through a circuit — for eight, twelve, twenty-four hours, with memory sampled throughout. A plot that climbs steadily and never comes down is a leak, and leaks found in QA are enormously cheaper than leaks found by players.
On mobile, thermals dominate. A phone will hit sixty frames per second for the first eight minutes and then throttle hard. Any performance measurement that does not account for sustained load is measuring the wrong thing, which is why a soak run on a warm device tells you more than a benchmark on a cold one.
This is the format you will be marked on throughout the course.
Performance and soak testing is one of eight disciplines covered in the 12-week course.
