Consica Labs

Consica Labs
Chapter 13

Heat, Power & Cooling

TDP, thermal management, and power efficiency

Introduction

Every time a CPU executes instructions, it consumes electrical power — and almost all of that power is converted into heat. A modern high-performance CPU can draw over 200 watts of power, equivalent to two bright incandescent light bulbs, all concentrated into a chip smaller than your thumbnail. Managing this heat is one of the greatest challenges in processor design.

The relationship between power, heat, and performance is governed by physics. Higher clock speeds and voltages increase power consumption quadratically — doubling the voltage quadruples the power draw. This is why CPU frequencies stopped scaling past about 4-5 GHz and why the industry shifted to multi-core designs instead. The power wall is the fundamental constraint that shapes modern CPU architecture.

Understanding heat and power is essential not just for CPU designers but for anyone building or maintaining computers. Proper cooling ensures your CPU can maintain its maximum performance without thermal throttling — the automatic reduction in clock speed when temperatures get too high. Good thermal management also extends component lifespan and reduces noise from cooling fans.

How It Works

A CPU is like a tiny electric heater that happens to do useful calculations. The electrical resistance of billions of transistors switching billions of times per second generates enormous heat. Without a cooling system, a CPU would reach temperatures of over 200°C in seconds and physically destroy itself.

Household Object Analogy

Think of your CPU as a stovetop burner set to high. The burner itself gets extremely hot (the CPU die). You put a pan on it (the heatsink) to absorb and spread the heat. You add oil (thermal paste) to fill gaps between the pan and the burner for better heat transfer. Then you turn on the exhaust fan (case fan) to move the hot air out of the kitchen. Without any of these, the kitchen becomes unbearably hot and you have to turn down the burner (thermal throttling).

Deeper Dive

TDP (Thermal Design Power) is the maximum amount of heat a cooling system must dissipate when the CPU is running at its base clock speed under a realistic maximum workload. TDP is measured in watts. A CPU with a 125W TDP requires a cooler capable of handling at least 125W of heat. TDP is not the same as power consumption — actual power draw can exceed TDP under heavy loads, especially when turbo boost is active.

TDP ranges vary enormously. Low-power mobile CPUs like the Intel Core i5-1235U have a TDP of just 15W, allowing them to run fanless in some designs. Desktop CPUs like the Intel Core i9-14900K have a TDP of 125W but can draw over 250W under full load with turbo boost. Server CPUs like the AMD EPYC 9654 have a TDP of 360W, requiring massive cooling solutions and significant data centre power budgets.

Thermal Throttling

Thermal throttling is the CPU's self-protection mechanism. When the internal temperature exceeds a safe threshold (typically around 95-105°C for modern CPUs), the processor automatically reduces its clock speed and voltage to lower heat generation. This prevents permanent damage but also reduces performance — sometimes dramatically.

Throttling can be caused by inadequate cooling (small heatsink, poor airflow, dusty fans), high ambient temperature, or extended heavy workloads. In laptops, thermal throttling is a common issue because the cooling system is constrained by size and weight. A laptop might run at full speed for only 30-60 seconds before throttling kicks in — this is why "sustained performance" benchmarks are more meaningful than burst performance.

Cooling Methods: Air vs. Liquid

Air cooling is the most common and cost-effective method. A metal heatsink (typically copper heat pipes bonded to aluminium fins) sits on top of the CPU, and a fan blows air across the fins to carry heat away. High-end air coolers like the Noctua NH-D15 can handle CPUs with TDPs up to 250W and are nearly silent in operation. Air cooling is reliable, requires no maintenance, and has no risk of leaks.

Liquid cooling (also called water cooling) uses a pump to circulate coolant through a block attached to the CPU and then through a radiator where fans expel the heat. AIO (All-In-One) liquid coolers are sealed units that require no maintenance. Custom loop liquid cooling allows enthusiasts to cool the CPU, GPU, and other components with a single system, often achieving lower temperatures than air cooling, especially under sustained loads.

Power Efficiency and DVFS

DVFS (Dynamic Voltage and Frequency Scaling) is the technology that allows a CPU to adjust its voltage and clock speed on the fly based on workload demand. When the CPU is idle, it can drop to a very low frequency and voltage, consuming just a few watts. When a demanding task starts, it ramps up within microseconds. This is why modern CPUs spend most of their time at low power states, contributing significantly to battery life in laptops and energy savings in data centres.

Undervolting is a technique used by enthusiasts to reduce CPU voltage while maintaining the same clock speed. This reduces power consumption and heat output, which can improve sustained performance by reducing thermal throttling. However, too little voltage causes instability and crashes. Undervolting requires careful testing and is not officially supported by most CPU manufacturers, though it is generally safe when done correctly.

Key Insight

Power and heat are the fundamental constraints on CPU performance. The laws of physics — not engineering — prevent CPUs from running at 10 GHz or having 1000 cores on a single chip. All modern CPU design is, at its core, a battle against heat. Every architectural innovation, from pipelining to multi-core to chiplet designs, is ultimately about doing more work within strict power and thermal budgets.

Advanced

At a deeper level, heat, power & cooling involves rules and patterns that engineers use worldwide. TDP follows standards so different brands and devices can still work together. That is why your phone, school laptop, and game console can all connect to the same network or use the same apps.

Thermal Throttling does not happen in a straight line. Systems often use backup paths, error checking, and retries so information arrives correctly. When something fails, smart Heatsink design helps the system recover instead of shutting down completely.

Scientists and engineers keep improving these systems every year — making them faster, safer, and more energy-efficient. The ideas you learn in this chapter are the same building blocks used in real data centers, robots, apps, and websites around the world.

Vocabulary Table

Term Definition
TDPThermal Design Power — the maximum heat a cooling system must dissipate under realistic maximum load, measured in watts
Thermal ThrottlingAutomatic reduction of clock speed when CPU temperature exceeds a safe threshold to prevent damage
HeatsinkA metal component (typically copper and aluminium) that absorbs and dissipates heat from the CPU
Liquid CoolingA cooling method using liquid coolant pumped through a block on the CPU to a radiator where heat is expelled
UndervoltingReducing CPU voltage while maintaining clock speed to lower power consumption and heat output
Power EfficiencyThe ratio of computational work performed to energy consumed, often measured in performance per watt
DVFSDynamic Voltage and Frequency Scaling — technology that adjusts CPU voltage and speed based on workload demand
Thermal PasteA thermally conductive compound applied between CPU and heatsink to fill microscopic gaps and improve heat transfer
Air CoolerA cooling solution using a metal heatsink and fan to transfer heat away from the CPU into the surrounding air
Fan CurveA configurable mapping of CPU temperature to fan speed, balancing cooling performance against noise

Fun Facts

The hottest consumer CPU ever made was the Intel Pentium 4 Extreme Edition (Prescott), which had a TDP of 130W in 2005 — a massive amount for its time. Today, that same TDP is considered mid-range, with flagship CPUs drawing over 250W.

Data centres are such massive consumers of electricity that they account for approximately 1-2% of global electricity use. A single large data centre can consume as much power as a small city. This is why efficiency is paramount in server CPU design.

Liquid metal thermal compounds like Thermal Grizzly Conductonaut can improve heat transfer by up to 10-15°C compared to standard thermal paste. However, they are electrically conductive and can destroy a CPU if spilled onto the circuit board — they are strictly for experienced enthusiasts.

Some extreme overclockers use liquid nitrogen (LN2) or dry ice to cool CPUs to sub-zero temperatures. At -196°C, LN2 cooling dramatically reduces electrical resistance, allowing clock speeds of 8 GHz or more — but the setup costs thousands and can only run for a few minutes before the LN2 evaporates.

Modern CPUs are so efficient at idle that they can run without any active cooling at low loads. Intel's NUC and Apple's MacBook Air with M-series chips can operate completely silently for light tasks because their fans only spin up under heavy load. This is achieved through aggressive DVFS and low-power core architectures.

Interactive Diagram

Launch the interactive diagram to explore Heat, Power & Cooling.

Open Interactive Diagram

The interactive diagram for this chapter demonstrates Heat and Power. It shows a thermal diagram of a CPU showing heat generation, power consumption, and cooling solutions.

What to explore:

  • increase CPU load; watch temperature rise; toggle different cooling solutions; see power draw change
  • CPUs generate significant heat under load — effective cooling is essential to maintain performance and prevent thermal throttling

Knowledge Check

1. What does TDP (Thermal Design Power) represent?

Answer: The maximum amount of heat a cooling system must dissipate under realistic maximum load

2. What happens when a CPU reaches its maximum safe temperature?

Answer: The CPU automatically reduces its clock speed (thermal throttling) to lower heat generation

3. What is DVFS (Dynamic Voltage and Frequency Scaling)?

Answer: A technology that adjusts CPU voltage and clock speed dynamically based on workload to save power