Power supplies, wattage, connectors and keeping the heat out

Objective 3.4 · Hardware · 25% of the exam

Why this matters

Power and cooling are the two systems that cause faults everywhere else. A marginal power supply produces symptoms that look like memory faults, storage faults, graphics faults and operating system corruption — and it produces them intermittently, which makes them the most expensive kind to chase.

This is the capstone lesson for the hardware domain because it is where the other components' requirements are added up. Choosing a supply is arithmetic over everything in domain 3, and cooling is the constraint that decides whether the processor from lesson 18 ever reaches its rated speed.

It is also where the safety rule from lesson 2 is most important. A power supply is not opened. Everything in this lesson is done from outside the case of the unit.

The lesson

Reading a power supply label: rails, wattage and what the headline number means

Every supply has a label stating what it can deliver on each rail — the separate voltages it produces.

  • +12 V does nearly all the work in a modern machine: processor, graphics, fans, drive motors. The 12 V figure is the one that matters.
  • +5 V and +3.3 V supply logic, older drive electronics and some board circuitry. Modern machines draw far less on these than machines from twenty years ago, which is why old supplies are poorly matched to new builds even when the total wattage looks adequate.
  • −12 V survives for legacy serial circuitry, at tiny current.
  • +5 V standby powers the board while the machine is "off", which is what allows the power button, wake-on-LAN and USB charging to work.

The headline wattage is the total, and it is only meaningful if the 12 V rail can supply enough of it. A 600 W supply with a weak 12 V rail is worse than a 500 W supply with a strong one.

Two more label facts:

  • Efficiency ratings (the 80 PLUS tiers) describe how much mains power is wasted as heat. A more efficient supply costs less to run and produces less heat, and does not deliver more power.
  • Single against multiple 12 V rails describes how the 12 V capacity is divided and protected. Either is fine; the total and the per-rail limits are what matter.

Connectors by name, and which ones a modern graphics card needs

  • 24-pin ATX main connector to the motherboard. Made of a 20-pin block and a 4-pin block, which is why older 20-pin supplies partially fit.
  • 4+4-pin EPS / CPU power at the top of the board. Modern boards may have two. A machine that does nothing when powered on frequently has this connector missing, because it is awkward to reach and easy to forget.
  • PCIe power, 6-pin and 6+2-pin, for graphics cards.
  • 12VHPWR / 12V-2x6, the newer high-power graphics connector carrying up to 600 W. It must be fully seated; partial insertion has caused connector damage and is a known failure mode.
  • SATA power, the flat keyed connector for drives.
  • Molex, the older four-pin connector, still used by some fans and accessories.
  • Berg, the small floppy connector, historical.

Modular supplies let unused cables be detached. The safety point: cables are not interchangeable between manufacturers, even where the connector at the supply end fits, because pinouts differ. Using the wrong modular cable can destroy components, and this is a real and repeated field failure.

Adapters — SATA to Molex, or doubling one PCIe connector into two — exist and should be treated with suspicion. An adapter cannot create capacity; it only moves the load somewhere that may not be rated for it.

Sizing a supply for a build without guessing, and the cost of oversizing

The method:

  1. Add the processor's thermal design power and the graphics card's board power. Together these are usually 70 to 90 per cent of the total.
  2. Add the rest — roughly 10 W per drive, a few watts per fan, and a small allowance for the board and memory.
  3. Add headroom, typically 20 to 30 per cent. Supplies are most efficient around half load, transient peaks exceed the average, and capacitors age.
  4. Check the 12 V rail specifically against the processor and graphics demand.
  5. Check the graphics card manufacturer's recommendation, which is a system-level figure accounting for transients.

Oversizing has a cost, and the exam may ask about it: a very large supply runs at a low fraction of capacity where efficiency is poorer, and it costs more. But undersizing is far worse, and its symptoms are the ones that waste days:

  • Shutdowns or reboots under load, with no error.
  • Instability that only appears when the graphics card is busy.
  • Random errors that look like memory or storage faults.
  • A machine that will not start at all with everything connected, and starts fine with the graphics card removed.

When several unrelated components appear to be failing at once, suspect the supply before replacing any of them.

Air cooling: airflow direction, dust, and why case fans are a system rather than parts

Air cooling is a path, not a collection of fans. Air enters at the front and bottom, passes over the components, and leaves at the rear and top. Fans fitted without regard to that path can actively make things worse.

The principles:

  • Front and bottom intake, rear and top exhaust. Heat rises; work with it.
  • Positive pressure — slightly more intake than exhaust — pushes air out through gaps rather than pulling dust in through them, so filtered intakes keep the inside cleaner.
  • Cable management is cooling. A bundle of cables across the front intake is a wall in the air path.
  • Blanking plates and drive bay covers keep the path intact; open slots let air short-circuit past the components that need it.
  • Fan direction is marked on the frame with arrows for airflow and rotation. Fitting one backwards is easy and produces a machine that cools worse after being "improved".

Dust is the slow failure. It insulates heatsinks, clogs filters and unbalances fans. The symptoms — gradually louder fans, gradually worse performance under load, eventual thermal shutdowns — arrive slowly enough that nobody dates the start. Cleaning is compressed air or a low-static blower, with fans held still while blowing, because spinning a fan backwards can generate enough voltage to damage the circuit it is connected to.

Component-level cooling to name: the processor heatsink and fan, the graphics card's own cooler, heatsinks on board chipsets and M.2 drives, and thermal paste or pads as the interface between a component and its heatsink.

Liquid cooling in scope for this exam, including what fails and how it is noticed

Liquid cooling moves heat to a radiator that has more surface area than a heatsink could have near the processor.

  • All-in-one units arrive sealed: a pump and cold plate, tubing, and a radiator with fans. No maintenance beyond cleaning and, in some designs, eventual replacement.
  • Custom loops use separate blocks, pump, reservoir, radiator and tubing. More capable, more expensive, and they require topping up and periodic cleaning.

Advantages: much greater heat capacity for sustained loads, and a quieter or smaller footprint near the processor. Disadvantages: more components that can fail, a pump that is a single point of failure, and the possibility of a leak.

Failure modes and their signatures:

  • Pump failure. Temperature rises very rapidly under load because a radiator with no flow is not cooling anything. The signature is a machine that overheats within seconds rather than minutes.
  • Air in the loop. Gurgling noise and inconsistent temperatures.
  • Permeation over years on sealed units, gradually reducing coolant.
  • Leaks, which are rare and serious. Power off immediately.
  • Radiator dust, which behaves exactly like heatsink dust.

Mounting orientation matters: the radiator should sit so that any air in the loop collects away from the pump. That is why sealed units are usually specified as radiator at the top or front rather than below the processor.

For the exam, the comparison is the examinable part: air cooling is simpler, cheaper and has fewer failure modes; liquid cooling handles more sustained heat in less space near the processor. Both are correct answers to different stated requirements.

Practise what you just read

1. Which rail does most of the work in a modern computer?

Select one

  1. The 5 volt rail, which supplies the logic on the board
  2. The 3.3 volt rail, which supplies the memory and chipset
  3. The 12 volt rail
  4. The 5 volt standby rail, because it remains energised whenever the machine is connected to the mains supply
Show answer

C. Processor, graphics, fans and drive motors all draw from 12 volts. That is why the 12 volt capacity matters more than the headline total when a supply is being assessed.

2. What does an efficiency rating on a power supply describe?

Select one

  1. The ratio between the capacity of the 12 volt rail and the total wattage the unit is able to deliver across all of its outputs
  2. The proportion of its rated output it can sustain continuously
  3. The accuracy with which it holds each rail within tolerance
  4. How much mains power is wasted as heat
Show answer

D. A more efficient supply costs less to run and produces less heat. It does not deliver more power, which is the misconception the rating invites.

3. Which connector is most often forgotten during a build?

Select one

  1. The CPU power connector at the top of the board
  2. The 24-pin main connector to the motherboard
  3. The PCI Express power connectors on the graphics card, which are awkward to reach once the card has been fitted into its slot
  4. The SATA power connector to the boot drive
Show answer

A. It is awkward to reach and easy to miss, and a machine without it does nothing at all when the power button is pressed. That symptom looks far more serious than it is.

8 more questions on this objective are part of the full course.

Practise the full question bank in the exam simulator

Hands-on labs

All hands-on labs

This is an independent study companion for CompTIA A+ Core 1 220-1201 and is not produced by or endorsed by CompTIA.