TDP vs Measured Power: Why Heat Sinks Get Undersized
TDP is a design target, not a measured power limit, and treating it as the real dissipation is why heat sinks get undersized: a chip rated at 65 W TDP can draw 45 W on an average workload, 88 W on a short boost burst, and every vendor defines the number differently. Size the thermal solution from measured worst-case sustained power plus margin — never from the TDP sticker alone.
The letters stand for Thermal Design Power, and the emphasis belongs on design. TDP tells you what the cooling system should handle for the chip to run at its rated frequency under the vendor's reference workload — but the vendor picks the workload, the ambient, and the boost behavior. Your product runs your workload in your enclosure at your ambient, and the only number that matters for your heat sink is the power your chip actually dissipates in your worst case.
What TDP Actually Means — and What It Does Not
Chip vendors use TDP as a cooling specification: the power the reference cooler must dissipate to keep the chip inside its thermal limit at rated operation. It is not the maximum electrical power the chip can draw. Modern processors are allowed to exceed TDP temporarily through boost and turbo mechanisms, and some server parts run sustained workloads above their nominal TDP when power limits are unlocked. Compounding the problem, "TDP" is interpreted differently across vendors and even across product lines from the same vendor.
| Label you see | What it usually means | Danger if used as real power |
|---|---|---|
| TDP (nominal) | Reference cooler design point | Real bursts run above it |
| Boost/PL2 power | Short-duration upper limit (often 1.2–1.5× TDP) | Sustained boosts heat the sink slowly |
| Measured max sustained | Your workload, your board, measured | This is the number to cool for |
| Package power limit (PBP/MTP) | Configurable by firmware | May be set above TDP in your BIOS |
Takeaway: read the fine print. If the platform allows boost states or an unlocked power limit, the thermal design target is whatever the firmware will actually allow sustained — which can sit 20–50% above the TDP sticker.
Why Undersizing Happens in Real Products
Undersized sinks rarely come from malicious corner-cutting. They come from three honest mistakes: using TDP as the design power, testing with a typical workload instead of the worst case, and testing at room temperature instead of the product's real ambient. Each mistake is small; stacked, they are how a "65 W cooler" ends up on a board that draws 88 W in a 45 °C enclosure.
| Design input | Typical assumption | Reality in the field |
|---|---|---|
| Power figure | TDP sticker | Measured max sustained + boost excursions |
| Workload | Vendor benchmark | Your application, worst module, worst data |
| Ambient | 25 °C lab | 40–60 °C enclosure, sunlight, altitude |
| TIM aging | Fresh grease | 2–5 °C degradation over years |
| Airflow | Fan free-air spec | Restricted inlets, filter dust, backpressure |
Takeaway: the fix is not paranoia — it is margin discipline. The classic industrial rule is to design the thermal solution for the measured worst-case sustained power and then add 15–30% margin, which absorbs boost, aging, and airflow degradation in one stroke instead of itemizing each one.
A Realistic Example: The Numbers Behind a "65 W" Chip
Take a mainstream processor rated 65 W TDP. In a typical office workload it draws 35–55 W. Under a heavy multi-core load with boost enabled, it pulls 88 W — 35% above TDP — for bursts of tens of seconds to minutes. If the cooler was sized for 65 W with no margin, the burst pushes the junction temperature up; whether that matters depends on the duty cycle, because the heat sink's thermal mass absorbs short bursts. The failure mode appears only when the burst is long enough or the workload is sustained: the sink slowly warms, the fan ramps, and the chip throttles in the middle of the user's render.
| Load condition | Typical chip power | What the sink sees |
|---|---|---|
| Idle/light office | 15–35 W | Easy; sink near ambient |
| Sustained typical workload | 40–60 W | Steady state; sink reaches equilibrium |
| Boost burst (seconds–minutes) | 80–100 W | Transient; thermal mass absorbs short bursts |
| Worst-case sustained (all cores, unlocked) | 80–100 W+ | The design case — sink must hold this |
Takeaway: short bursts are handled by thermal mass, not fin area — a heavy finned sink can absorb a 30-second boost with only a few degrees of movement. The sink must be sized for the longest sustained worst case, and the fan curve must be set so the product reaches equilibrium below the throttle point, not after it.
How to Find the Real Number
Measure it. On the bench, run the product at maximum load with a power meter or current probe on the processor's supply rail while logging temperature, and let it run long enough to reach equilibrium — typically 30–60 minutes, not the 5 minutes most lab tests use. Repeat at the worst ambient the product will see, with the enclosure closed and the fan inlet as restricted as it will be in the field. The sustained measured power at that point is your design number. If measurement is impossible before the mechanical design freezes, use the vendor's maximum sustained figure — the boost or unlocked limit, not the TDP — and apply the margin.
Specifying the Sink with the Right Number
Once the real power is known, the sizing math is the same as any other thermal budget: required sink resistance equals the allowed temperature rise divided by the measured power, minus the interface. The heat sink sizing calculation guide runs the full worked example. What changes with measured power is the confidence: a sink sized from a measured 88 W sustained with 20% margin is a released design; a sink sized from a 65 W sticker is a field-failure report waiting for a hot summer. For the mechanical side, the construction choice follows quantity and geometry as always — extruded heat sinks for volume profiles, CNC-machined heat sinks for low-volume or complex bases — and the same rule applies to power electronics, where IGBT and SiC module losses are similarly undersold by datasheet typicals; our IGBT power module cooling guide shows the module-side numbers. When you send the inquiry, include the measured power and the ambient, and the factory can confirm the sink rather than guess at it.
Email sc@bquq.com or WhatsApp +86 137 1315 7787 with your PDF/DXF/STEP file. An engineer reviews it and replies with price, lead time and DFM notes on working days.
Frequently Asked Questions
Q: Is TDP the same as maximum power consumption?
A: No. TDP is the cooling design point the vendor specifies for a reference workload. Real chips can draw less on typical loads and more during boost bursts or unlocked sustained operation — often 20–50% above TDP. Measure your worst case instead of trusting the sticker.
Q: How much margin should I add over measured power for a heat sink?
A: Use 15–30% over measured worst-case sustained power. That margin absorbs boost excursions, TIM aging, dust on the fan inlet, and ambient variation without itemizing each effect. Short bursts need less margin because the sink's thermal mass absorbs them.
Q: Why does my processor throttle even though the cooler is rated above TDP?
A: Either your workload sustains power above the cooler's rating, the ambient is hotter than the rating assumed, or airflow is restricted by the enclosure. Log actual power and temperature together — the pair will show which limit is being hit and when.
Q: Does a bigger heat sink help with short power bursts?
A: Only up to a point. Fins add surface for steady-state heat rejection; thermal mass (mostly base and fin metal) absorbs transients. A heavy sink smooths short bursts, but a sustained overload will always win eventually — size for the sustained case and let mass handle the spikes.
Q: How long should a thermal test run to catch undersizing?
A: Long enough to reach equilibrium at worst-case load — typically 30–60 minutes, sometimes longer for large passive sinks with high thermal mass. Short tests miss slow warming and return optimistic numbers that disappear in the field on a hot afternoon.
Related Articles
- heat-sink-sizing-calculation — More from the BQUQ Thermal Management engineering series.
- igbt-power-module-cooling-guide — More from the BQUQ Thermal Management engineering series.
- heat-sink-selection-workflow — More from the BQUQ Thermal Management engineering series.
Data Sources and Verification
Tolerances, cycle times and price ranges in this guide come from BQUQ production records at our Dongguan plant, where CNC machining (±0.005 mm), stamping, custom springs and heat sinks run under one roof. BQUQ is an ISO 9001:2015 certified factory; the certificate and batch inspection reports are available on request with every quotation.
Related Resources
- About BQUQ: an ISO9001-certified source factory in Dongguan running four production lines under one roof.
- Heat sinks and thermal parts: extruded, CNC-machined and stamped options from the thermal line — extruded heat sinks, CNC-machined heat sinks, stamped heat sinks.
- Industry trends: manufacturing, material market, and sourcing analysis for buyers.
- Technical articles: engineering guides and process comparisons — more where this article came from.
- FAQ hub: quick answers on CNC, stamping, springs, and heat sinks.
- Case studies: real parts and real numbers from projects we engineered and delivered.
- Contact us: send your drawing and get a quote within 12 working hours.
Authored by the BQUQ Engineering Team. BQUQ is an ISO9001-certified source factory in Dongguan, China, running CNC machining, metal stamping, custom springs, heat sink and collet lines under one roof. Send drawings to sc@bquq.com or WhatsApp +86 13713157787 for a quote within 12 working hours. www.bquq.com
