Thermal Design Power
Thermal design power (TDP) is the amount of heat a chip's cooling system has to be built to remove, not the maximum power the chip can ever draw — the two numbers get confused constantly, and the gap between them is exactly why a cooler that matches a chip's TDP on paper can still let that chip throttle under a real workload. A vendor picks TDP as a sustained, representative power level the cooling solution must handle continuously; a chip's actual instantaneous power draw can spike well above that number for short bursts, because silicon has thermal mass and a brief spike doesn't raise junction temperature nearly as much as the same power sustained for minutes would.
Modern chips actively manage this gap with two numbers instead of one, not by accident. Intel's PL1 and PL2 power limits and AMD's PPT (package power tracking) formalize exactly the distinction a single TDP number used to blur — PL1 (or its equivalent) tracks close to the sustained cooling-system target, while PL2 allows a much higher power ceiling for a bounded time window, letting a chip turbo aggressively right after a workload starts before settling down to whatever the cooling solution can actually sustain indefinitely; a laptop and a desktop using the identical silicon can ship with very different PL2 durations and ceilings purely because their chassis cooling capacity differs.
A cooling solution that is correctly sized to TDP is not under-built if a chip spikes above it briefly. The entire point of separating a sustained target from a burst ceiling is to let short, high-power bursts happen without demanding a cooler built for peak power continuously, which would be wildly oversized, more expensive, and heavier than the workload actually requires most of the time — a laptop chassis built to a 28-watt TDP routinely lets its chip pull 60 or more watts for a handful of seconds at the start of a task specifically because that's cheap thermal headroom the system already has, not a design flaw.
TDP numbers from different vendors have never meant exactly the same thing, which makes cross-vendor comparison genuinely tricky. Intel, AMD, and Arm-based SoC vendors have each defined their sustained-power metric slightly differently over the years — some measurement methodologies emphasize base clock sustained power, others emphasize a specific thermal/electrical operating point — so a published TDP number is really only directly comparable within one vendor's own product line across generations, and comparing TDP figures across vendors as if they were the same measurement has been a recurring source of confusion in product reviews and procurement decisions alike.
Read TDP through a *cooling-budget* lens rather than a *power-ceiling* lens — it was never meant to describe the most power a chip could draw, it describes the heat load engineers decided the cooling solution has to be able to carry all day, every day, with everything above that line being a deliberate, time-limited exception rather than the number failing at its job.