Compute tuning / experimental playground
Find your operating point.
More clock is only half the story.
Make useful work fit inside a finite budget.
Throughput trial
A deterministic toy model. No device measurements, hardware controls or real benchmarks.
C / Run complete
Your operating point
Save this run to your browser notebook.
D / Your browser, your experiments
Keep the useful mistakes.
Select two saved runs to compare. Rankings are local and specific to each challenge.
| Compare | Experiment / challenge | Inputs | Work | Peak °C | Units/J | Goal | Actions |
|---|---|---|---|---|---|---|---|
| No saved experiments yet. Run a challenge and keep your first operating point. | |||||||
Local challenge records +
The model is open
Simple enough to question.
Sixty one-second steps. A 24°C starting point. Useful output falls when power is limited or heat becomes unstable.
Inspect the equations +
Inputs: f = clock in GHz; B = total power budget in W; c = cooling from 0 to 100.
Demand D = 12 + 8f² W
Cooling draw F = 2 + 0.12c W
Power factor p = min(1, (B − F) / D)
Throttle q = clamp(1 − max(0, T − 75) / 40, 0.35, 1)
Compute draw P = Dpq
Temperature step ΔT = (P − (0.25 + 0.012c)(T − 24)) / 20
Stability s = clamp(1 − max(0, T − 80) / 50, 0, 1)
Useful rate = 100fpqs units/s
Energy sums (P + F) over 60 seconds. Work sums useful rate. Efficiency = work / energy. Output at each step uses temperature before that step; the chart shows temperature after it. These equations are educational assumptions, not hardware forecasts.
Proposed application credits
A wider experiment shelf.
Clockmoth Credits are planned for scenario packs, experiment recipes and team leaderboard tools. The local lab is free; no token is needed.