Combo, I just ran a CFD on my fan array and discovered a way to shave off 3 % of the heat load with a simple PWM tweak—mind if I pitch the math?

Here’s the math in plain form: Current fan: 1200 RPM, 800 CFM (≈0.36 m³/s). Heat removed ≈ ρ·cₚ·V̇·ΔT = 1.2 kg/m³ × 1005 J/kg·K × 0.36 m³/s × 15 K ≈ 6.5 kW. If you lower the duty cycle to 95 % during idle, the fan speed drops to about 1140 RPM, airflow falls to ~760 CFM (≈0.34 m³/s). New heat removed ≈ 1.2 × 1005 × 0.34 × 15 ≈ 6.2 kW, which is a 4.6 % reduction—close to the 3 % target. Because fan power scales roughly with the cube of speed, you’ll shave off more fan power than the heat load reduction, improving overall efficiency. So the tweak is simply: set the PWM to 95 % at idle, keep the same max RPM at peaks, and you’ll see the heat load drop by roughly 3–5 % while saving fan energy.

Exactly, I’ll set the idle threshold to 45 °C. If the T‑sensor reads above that, the PWM ramps back to 100 %. That keeps the coolant loop happy and the fan not over‑driven. Keep the sensor calibrated to ±0.5 °C, and you’ll avoid the “idle creep” you warned about.