Week 04 · lesson
Lesson 2: Cooling Is Part of Reliability
Core path: 42 minutes
Heat is not a side issue.
A computer that cannot move heat away from its components cannot sustain reliable operation.
Where the heat comes from
CPUs, GPUs, voltage-regulation components, storage devices, and power circuitry all produce heat while operating.
The cooling system has to move that heat from the component into the surrounding environment.
A useful path is:
chip
→ thermal interface
→ heatsink / cold plate
→ moving air or liquid loop
→ case airflow
→ room
Every boundary matters.
Heatsinks and thermal interface material
A heatsink provides surface area so heat can move into surrounding air. Thermal paste or another approved interface material fills microscopic gaps between mating surfaces.
More paste is not automatically better. The goal is a correct interface, not a frosting layer.
Air cooling
Air-cooled systems commonly use:
- CPU heatsink and fan;
- GPU cooling assembly;
- intake fans;
- exhaust fans;
- unobstructed vents.
Airflow should have a path. A case packed with fans can still cool badly when cables, dust, orientation, or blocked vents disrupt that path.
Liquid cooling
Closed-loop AIO systems move heat using a pump, liquid, radiator, and fans. They add components and failure modes, so they are not automatically the correct answer for every PC.
A+ symptom set: thermal and power clues
Several symptoms can point toward the power/cooling boundary:
- overheating;
- random shutdown under load;
- sluggish performance caused by thermal throttling or resource stress;
- application crashes during unstable thermal/power conditions;
- unusual noise from a failing/obstructed fan or pump;
- burning smell from unsafe electrical/thermal failure.
The last one is a stop-work condition. Do not keep powering the system to reproduce a burning smell.
Guided case: shuts down during gaming
Evidence:
- system boots normally;
- light desktop work is stable;
- fans become loud under load;
- performance drops;
- system eventually shuts down;
- after cooling, it starts again.
This pattern makes a thermal problem a strong hypothesis.
It does not prove thermal paste is the cause.
Possible boundaries include:
- failed or obstructed fan;
- dust-loaded heatsink;
- poor cooler mounting;
- blocked airflow;
- pump failure on a liquid cooler;
- excessive ambient temperature;
- another hardware or PSU problem that appears under load.
Guided case: unusual noise plus rising temperature
A system begins making a rattling/grinding fan-area noise. CPU temperature rises and fan RPM becomes inconsistent.
That evidence justifies inspecting the fan/cooler path with the system safely powered down. It does not justify opening the PSU or touching a spinning fan.
Student action: thermal evidence table
For each symptom, write the strongest next check:
| Symptom | Next check | What it could prove |
|---|---|---|
| CPU fan not spinning | inspect fan connection/obstruction with power safely removed | whether the fan path is connected/free |
| vents packed with dust | inspect and clean under approved procedure | whether airflow restriction is present |
| high temperature only under load | compare temperature/fan behavior at idle vs load | whether heat removal fails as demand rises |
| random shutdown with high temperature | inspect cooling plus PSU/load evidence | whether thermal/power instability correlates with load |
| burning smell | stop power and escalate | unsafe hardware condition requiring service |
| shutdown with normal temperatures | investigate other power/hardware evidence | thermal hypothesis becomes weaker |
Evidence checkpoint
A technician does not diagnose "overheating" because a computer feels warm. A useful diagnosis connects workload, temperature or fan evidence, airflow, repeatable symptoms, and safety boundaries.
Visual reference
See the system before you troubleshoot it
Form factor is only one clue. Interface, power, slot, and workload requirements determine whether a component belongs in a system.
Use it for: Compare storage interfaces and trace power, expansion, and display dependencies.
Read it. Prove it.