MR_

POLAND, 2026

Contact
Menu

LAB / REAL-TIME / SIMULATION

Fixed Timestep

Explore the relationship between frame duration, simulation steps and work carried into the next frame.

DEMONSTRATION / DETERMINISTIC MODEL

Time (ms)048Executed48Deferred0Remainder2
Steps executed
3
Whole steps deferred
0 ms
Fractional remainder
2 ms

One frame, starting with an empty accumulator. Whole steps beyond the cap are deferred. The fractional remainder also carries over. No actual frame timing is measured.

Question

How much simulation work fits into one frame when the number of steps is capped?

Setup

The accumulator begins empty. The selected frame duration is divided into whole simulation steps and a fractional remainder. Only the allowed number of whole steps executes. Remaining time carries into a later frame.

Interpretation

The step count is arithmetic, not a measurement of how expensive a simulation step is. Capping work can avoid an unbounded catch-up loop, but deferred time must be handled deliberately by the application. This model does not silently discard it.

Limitation

There is no rendering, interpolation, wall-clock sampling or actual game loop here. The model isolates one scheduling decision so its consequences are visible.

Methodology & reproducibility

Evidence type
Deterministic model, not a runtime benchmark.
Inputs
The values shown in the controls above. Keep them identical when comparing results.
Hardware & trials
No hardware measurements or repeated timing samples are collected.
Median / p95
Not applicable to calculated outputs. These require a distribution of measured samples.

Use this model to form a hypothesis. To validate it in a project, record the source revision, build configuration, hardware, workload, warm-up and trial count; then compare baseline and changed builds under the same conditions.

← All experiments