Naquadah Fuel Refinery: perfect overclocks (×4 power, ×4 speed — EU per liter never changes), capped at coil tier − recipe tier. No parallels — one recipe at a time. Unlimited tier skips; energy is the sum of all hatches (regular, multi-amp, laser all count), so any supply that fits the ×4 steps works.
Liters per second sweeping total energy amps (log scale), one line per coil tier. Every ×4 supply = one more perfect overclock (×4 speed) until the coil cap — then the line goes flat: extra power buys nothing. Marker = current setting.
Actual EU/t consumed. Perfect overclocks quadruple draw and speed together, so the staircase tracks output exactly — anything above the last step is wasted supply headroom.
How NFR computes this: recipe requires coil tier ≥ recipe tier (T1–T4 field restriction coils). Overclocks are perfect (enablePerfectOverclock()): each needs ×4 total energy input and gives ×4 speed at ×4 draw — EU per recipe is invariant, so overclocking is pure win. OC count = min(coil tier − recipe tier, ⌊log₄(supply ÷ recipe EU/t)⌋), duration floor 1 tick. Power model: setAvailableVoltage(getMaxInputEu()) — the summed EU/t of every energy hatch (regular 2A each, TecTech multi-amp and laser hatches allowed) — with amperage 1 and unlimited tier skips, so hatch tier is irrelevant, only total EU/t matters. Max parallel is 1 and durations never reach sub-tick with the coil cap (≤3 extra OCs), so the NFR never parallelizes — for more throughput past the coil cap, build more refineries. Batch mode supported (logistics only, no rate change).