PC hardware, explained: power, fans, RAM, storage & everything you can touch

The single most useful way to think about hardware is as a dependency chain, not a parts list. The power supply converts wall power into the low voltages the components use. The motherboard is the connective tissue that everything plugs into. The CPU does the general-purpose work and needs both power and cooling to do it. RAM gives the CPU its fast workspace. Storage holds the operating system, the applications, and your files. The case and its fans move air across all of it so nothing overheats. If you hold that picture in your head, most faults become a question of "which link in the chain is the one that's failing?"

Why faults show up in the wrong place

A computer doesn't label its failures for you. The power supply is the classic example: when it can't deliver clean, steady power, the symptom is often something that looks like a different part — a random shutdown under load, a graphics card that "crashes," or a system that simply won't post. A failing disk can make the whole operating system look broken. A RAM stick that's not fully seated can look like a dead motherboard. This is why hardware troubleshooting is about isolating variables, not about memorizing which part goes with which symptom.

The good news is that the chain also makes problems easier to isolate, because you can test one link at a time with a known-good part. Swap the RAM, then the GPU, then the power supply, and the fault reveals itself by elimination. That method — one variable at a time, safest first — runs through the whole of this guide and is the same one professionals use.

The power supply unit (PSU) converts the high-voltage AC from the wall into the low-voltage DC that the components actually use — mainly 12V, with 5V and 3.3V rails for other duties. It's the one component that, if it fails, can make the entire machine look dead, and it's the one you never open.

What the PSU actually does

Think of it as a translator and a gatekeeper. It steps the wall voltage down to the safe rails, and it regulates them so they stay steady under whatever load the system throws at it. Most of a modern system's power draw — the CPU and especially the graphics card — comes from the 12V rail. If that rail can't hold up, you get instability, crashes, or a machine that simply refuses to start. That's why the power supply is such a common hidden cause of problems that look like they belong to other parts.

Wattage: how much do you need?

Wattage is the PSU's maximum output. The goal is to cover the peak your system will actually draw, with a margin. The big consumers are the CPU (its TDP or peak power) and the graphics card (its spec'd power draw, which is usually the dominant number). Add those together, add a modest amount for the rest of the board, drives, and fans, and leave headroom — roughly 20–40% above your estimated peak is a common, sensible target. You want to avoid two extremes: an underpowered unit that's running at its limit, and a wildly oversized one that's a poor-quality bargain. If you're adding a powerful GPU, use the wattage that card recommends as your baseline.

Efficiency ratings

The "80 PLUS" tiers (White, Bronze, Silver, Gold, Platinum, Titanium) describe efficiency — how much of the wall power actually reaches the components versus being wasted as heat. A higher tier means less waste, which is gentler on the power supply and slightly better on the electricity bill. It is not, by itself, a measure of safety or build quality, so don't treat a gold rating as a guarantee. For most systems, a quality unit in the bronze-to-gold range from a reputable maker is a fine, balanced choice.

Modular versus non-modular

A non-modular PSU has all its cables permanently attached. A modular one lets you plug in only the cables you use — full modular means every cable detaches, semi-modular means the essential ones (motherboard, CPU) are fixed and the rest are optional. Modular makes cable management cleaner and improves airflow, at a slightly higher price. The trade-off is worth it on most builds; on a budget machine, a solid non-modular unit is perfectly fine.

The connectors you'll meet

You'll handle a handful of PSU connectors, and learning to tell them apart saves a lot of confusion during a build. The shapes and labels are your friends — the CPU connector, for instance, can look similar to the 24-pin but is a different, smaller plug and is not interchangeable with it.

When the power supply is the suspect

The telltale signs: a system that powers on fine at idle but shuts down or reboots under load; a machine that won't start at all even though the outlet and cable are good; or parts that work in another machine but not this one. The power supply is also the hardest part to test conclusively — the definitive test is a known-good PSU (or a dedicated PSU tester). If everything else checks out and the symptoms track with how hard the system is working, the PSU is the leading suspect. It's a replace-it part, which is exactly as it should be.

• Sizing on the GPU alone — the CPU, drives, and fans add up too; use the whole system's draw.
• Reusing an old, underpowered unit with a new, much more powerful graphics card.
• Daisy-chaining one PCIe cable to power a card that wants two separate connectors — use separate cables from the PSU where the card needs more than one.

Heat is the quiet, long-term enemy of a PC. Good cooling isn't about making fans spin as fast as possible; it's about moving heat out of the case in a consistent way so every component stays within a safe operating range. A system that's just a few degrees cooler lasts longer, runs quieter, and is less likely to throttle or shut down under load.

How airflow actually works

Airflow is about a path, not a number of fans. Cool air is drawn in through intake fans (typically the front and bottom), passes across the heat-producing components, and warm air is pushed out through exhaust fans (typically the top and rear). What you want is a steady front-to-back or bottom-to-top flow, not a swirl of air going nowhere. More fans doesn't automatically mean better cooling — a well-arranged few fans beats a chaotic many.

Positive and negative pressure

If a case has slightly more intake than exhaust, it's under positive pressure — air is pushed out through the gaps and mesh, which tends to keep dust out. If it has more exhaust than intake, it's under negative pressure — air is pulled in through the gaps, which keeps temperatures a touch lower but lets more dust in. Either is fine as long as the flow is consistent; the choice is a trade-off between cleanliness and raw cooling. Most modern cases are designed around a light positive pressure with good, directed airflow.

Case fans: placement and which way they blow

The default layout that works for almost every case: one or two intake fans at the front, an exhaust fan at the rear, and (if there's a mount) one or two at the top for exhaust. To tell which way a fan blows, look at the frame — the side with the wiring and the little label sticker is the exhaust side (it blows air toward you), and the smooth, slightly concave side with the airflow arrow is the intake side (it draws air in). Fans are reversible in orientation just by mounting them the other way; they're not reversible in the wiring.

CPU coolers

The CPU runs hot because it's small and dense, so it gets dedicated cooling. An air cooler is a metal block with heat pipes and a fan that pulls heat off the processor and out through the case — simple, cheap, and extremely reliable. An AIO liquid cooler uses a pump and a cold plate on the CPU, pipes to a radiator, and fans on the radiator; it often cools better and can run quieter, at the cost of more parts and more money. Either works well for most systems; the choice is mostly about space, budget, and preference. Whichever you use, the cooler has to be firmly and evenly seated — a cooler that isn't properly mounted is the fastest way to overheat a CPU.

Thermal paste

Thermal paste fills the microscopic air gaps between the cooler's base and the top of the processor (air is a terrible conductor of heat, so those tiny gaps matter). It degrades slowly over years, and it gets ruined if you fully remove and replace the cooler. When you do a repaste, wipe off the old paste, apply a small amount centered on the processor (a pea-sized dot is a fine rule of thumb), and let the mounting pressure spread it. You don't need to repaste on a simple reseat — only when you've fully broken the contact.

Dust

Dust is cumulative and it's where it's at: on fan blades, inside heatsinks, and over the power supply's intake filter. A layer of dust acts as insulation — it makes the same fans work harder for the same cooling, so a dusty PC runs hotter and louder than the same PC clean. How much is "too much" is environment-dependent (pets and ground floors collect more), but if you can clearly see a fuzzy coating on the fan blades and fins, it's time for a clean. See the cleaning section for how to do it safely.

Fan noise

A fan that gets louder under load is doing its job — that's normal. What's not normal is a grinding, clicking, or whining sound, which usually means a bearing is wearing out. A fan that's loud even at idle points at dust, a blocked airflow path, or a failing bearing. Most fans are speed-controlled (either by the motherboard's PWM signal or a voltage-controlled DC fan), so a fan that's pinned at full speed at idle can also be a sign it's compensating for poor cooling — worth checking the temperatures and the dust.

What temperatures are "normal"

Exact numbers vary a lot by component, cooling, and room temperature, so treat these as rough guideposts rather than hard limits. A CPU that's comfortably in the 30s–50s at idle and the 60s–80s under heavy load is behaving normally; sustained operation near or above 90°C is a sign to look at the cooling. Graphics cards behave similarly, with sustained 85°C+ worth investigating. If a temperature was fine and then jumped, suspect the thermal paste, a fan that's no longer spinning, or a dust build-up.

RAM is the fast, temporary workspace the CPU reads and writes to constantly while it works. It's one of the easiest parts to install, one of the cheapest meaningful upgrades, and one of the most common causes of "my PC won't boot" and random instability. It's also the part most often confused with storage, so let's be clear about the difference first.

What RAM is (and isn't)

RAM is volatile — it holds data only while it's powered on, and it's wiped the moment the machine is off. That's why your open work isn't still there after a power cut. Storage (a hard drive or SSD) is the opposite: non-volatile, slower, and where everything permanently lives. When you "run out of memory," that's RAM; when you "run out of space," that's storage. A lot of "my computer is slow" comes down to not enough RAM, and a lot of "my computer is full" comes down to not enough storage — they're different fixes.

DDR generations

RAM comes in generations — DDR3, DDR4, DDR5 — each faster than the last and physically keyed differently, so you cannot mix generations in the same slots. The generation your board supports is fixed, and it's dictated by the CPU and motherboard together. Before buying any RAM, the first thing to confirm is which generation the system uses. Within a generation, speed increases are gradual, and a faster stick won't turn a DDR4 board into a DDR5 one.

Capacity, speed, and timings

Capacity is how much (in gigabytes) — and for most people this is the number that matters most; going from 8GB to 16GB is often a far bigger real-world improvement than any speed bump. Speed (in MT/s or "MHz") is how fast it transfers data. Timings (like "CL16") are the latency, and they're a secondary consideration for most users. A practical priority order: enough capacity first, then speed, then timings.

Single-channel versus dual-channel

Most motherboards have four RAM slots, paired so that putting sticks in the right two (often the outer pair — slots 2 and 4 — but check the manual) enables dual-channel mode, which can roughly double memory bandwidth. Running two matched sticks in the correct slots is almost always faster than a single stick of the same total capacity, even if that means two smaller sticks rather than one big one. If a board has two slots, populating both is the whole story; on four-slot boards, the manual's recommended slots matter.

XMP and EXPO

Most RAM is sold to run faster than the base "JEDEC" speed that the system uses by default. XMP (on Intel systems, and now broadly supported) and EXPO (AMD's equivalent) are saved profiles on the stick that tell the motherboard to run the RAM at its rated speed and timings. You enable the profile in the BIOS/UEFI. The catch: it's not guaranteed stable on every board, so if you enable it and the system becomes unstable, the first thing to try is disabling it and running at the base speed. It's an easy on/off, which makes it a good first thing to check when a new RAM kit misbehaves.

Installing RAM

1. Power off and unplug the machine, and drain residual charge (press the power button once).
2. Locate the RAM slots — they run along the side of the CPU socket, each with a clip at one or both ends.
3. Open the clips on the slot(s) you're using (the correct pair for dual-channel, per the manual).
4. Align the notch on the stick with the key in the slot, and press straight down, evenly, until the clips snap back into place. You should feel a definite click on each side.
5. Test with one stick first if you're troubleshooting, then add the second.

Troubleshooting RAM

RAM problems present as a machine that won't boot, beep codes, or random BSODs and instability. The diagnostic approach is the same every time: test one stick at a time, in a known-good slot. If a stick works in slot A but not slot B, the problem is the slot/board, not the stick. If a stick that works in another machine doesn't work here, suspect the board. And if you just enabled XMP/EXPO and things went unstable, disable it and retest at the base speed before assuming the stick is bad. A RAM problem is one of the cheapest and fastest things to rule out, which is why it's near the top of most "won't boot" checks.

Storage is where everything lives — the operating system, your applications, and your files. It's the part most people upgrade first, and its failure modes are very different between spinning drives and solid-state drives, so it's worth understanding the distinction before you buy or replace anything.

Hard drives versus SSDs

A hard disk drive (HDD) stores data on spinning magnetic platters read by a moving arm — it's fast enough for bulk storage, cheap per gigabyte, but slow to respond and fragile if moved while spinning, and it makes noise. A solid-state drive (SSD) stores data in flash memory with no moving parts — it's dramatically faster to respond, silent, and more shock-resistant, but pricier per gigabyte. For the drive that runs the operating system, an SSD makes the whole machine feel noticeably faster, which is why the HDD-to-SSD upgrade is the classic "everything got snappier" win.

SATA SSDs versus NVMe SSDs

Both are flash, but they connect differently. A SATA SSD is a 2.5" drive that connects to the board with a SATA data cable and a SATA power cable — it's limited by the SATA interface to roughly 500–550 MB/s. An NVMe SSD slots directly into an M.2 connector on the board and talks to the CPU over the much faster PCIe lanes, commonly in the thousands of MB/s. The key terminology trap: M.2 is the physical form factor (the slot and the stick shape), and NVMe is the fast protocol — most M.2 SSDs today are NVMe, but some older M.2 drives use the slower SATA interface, so the shape doesn't always tell you the speed. Check the drive's spec to be sure.

Interfaces and form factors

The common ones: a 3.5" HDD (the large desktop drive), a 2.5" SATA SSD (the laptop-sized drive), and an M.2 2280 NVMe SSD (the small stick, 22mm wide by 80mm long, that slots straight into the board). Before installing, confirm your board actually has the slot or port you need — not every board has a free M.2 slot, and using a particular M.2 slot can, on some boards, disable one or more SATA ports. The board's manual documents this, which brings us to the recurring theme in hardware work: the manual is your friend.

Installing storage

A SATA drive needs two cables: a SATA data cable (thin, flat, L-shaped ends) from the drive to a SATA port on the board, and a SATA power cable from the power supply. Mount it in a drive bay or bracket and you're done. An M.2 SSD is simpler on cabling — you angle it into the M.2 slot and secure it with a small screw (sometimes a stand-off is needed first), and the slot supplies both power and data. If the M.2 slot has a thermal pad or heatsink, make sure it's in contact. Either way, seating the connection fully is the step people most often skip and then wonder why the drive isn't detected.

Drive health

A failing drive announces itself in different ways by type. A failing HDD often starts making clicking or grinding noises, gets slower, or surfaces read errors — clicking is a strong mechanical-failure sign, and the right response is to back up immediately and plan to replace it. An SSD has no moving parts, so it fails less dramatically, but it does have a finite amount of data it can write over its lifetime (in practice, years of headroom for most people) and it can still show errors. Both types can report their health inside the operating system (drive utilities and the Event Viewer show errors), and a sudden "drive not detected" can be a dead drive, a dead cable, or a dead port — test the drive elsewhere to tell which. If a drive has gone missing, the hard drive not detected guide walks through isolating the cause.

Upgrading storage

The most common and rewarding storage upgrade is moving the operating system from an HDD to an SSD. You have two routes: clone the existing drive to the new SSD (keeping your current setup, files, and applications) or do a fresh install of the operating system onto the new SSD (cleaner, but you reinstall and restore). Either way, keep the old drive — as a backup or as cheap bulk storage. If slowness is the driver for the upgrade, the slow PC guide covers confirming that storage is actually the bottleneck first.

The motherboard is the board everything plugs into. It determines what's compatible (which CPU, which RAM, how much of what), and it's where all the physical connections come together. Understanding where things go on the board is half the battle in any build or repair.

What lives on the board

In one place: the CPU socket, the RAM slots, the PCIe slots (for the graphics card and add-in cards), the M.2 slots (for NVMe SSDs), the SATA ports (for SATA drives), the 24-pin ATX power connector, the CPU power connector, and a cluster of small pin headers for the front panel (power button, LEDs) and front USB ports. If you can point to each of these, you can build or service almost any desktop.

The CPU socket

The socket is where the CPU sits, and it's the most delicate thing on the board. There are two families: LGA (used by Intel, where the pins are on the board and the CPU has a flat contact plate) and PGA (used by AMD, where the pins are on the CPU). In both cases the golden rule is the same: never touch the pins. On an LGA board the pins are tiny and easily bent; on a PGA CPU the pins are on the chip you're holding. You lift the socket's retention mechanism, drop the CPU in aligned by its marker (a triangle or a notch), and lower the mechanism — no force. If it's going in, it goes in; if it's fighting you, it's misaligned.

PCIe slots

PCIe slots are where the graphics card and other add-in cards (and sometimes M.2 drives) go. They come in different lengths and "lanes" (x16 is the full-width slot a GPU uses; shorter x1/x4 slots serve smaller cards). The longest slot, usually the top one, is almost always where the graphics card goes — it's closest to the CPU, which is where it needs to be for speed. When installing a card, the slot has a latch that holds it, and a screw or bracket fixes it to the case.

SATA and M.2 ports

SATA ports (usually a row of small L-shaped connectors) are where SATA drives connect their data cable. M.2 slots are the small connectors where M.2 SSDs sit. Two things to know: an M.2 slot has a keying that must match the drive (the "M key" is common for NVMe), and — as noted in the storage section — using a particular M.2 slot can disable specific SATA ports on some boards. The manual lists which, so check it before you plan your storage layout.

Front-panel and USB headers

The little pin headers for the power button, reset button, hard-drive and power LEDs, and the front USB ports are the part of a build that trips up the most people, and they're easier than they look. Each has a labeled arrangement of pins with one pin missing (so the connector can only go on one way). The manual shows the exact layout for that board. Getting the power button header right is the difference between a machine that starts and one that does absolutely nothing when you press the button — which is why it's a standard "won't turn on" check.

Reading your manual

Here it is again because it's true: the motherboard manual is the single most useful reference in any build. It shows the socket, the RAM slot population for dual-channel, every header's pinout, which SATA ports a given M.2 slot disables, and the form factors the board supports. If you're not sure where something goes, the answer is in the manual — not in a guess. It's the one "reference" that's guaranteed to be correct for the exact board in your hands.

The CPU is where the general-purpose computing happens. It's small, dense, and powerful, which means it runs hot and needs dedicated cooling to do its job. It's also the most delicate part to install, so it's worth doing it right the first time.

What the CPU is (and isn't)

The CPU handles the broad, general work — running the operating system, launching applications, doing the everyday logic. It is not the graphics card: the GPU handles the specialized work of rendering images and video. A lot of confusion comes from mixing those up. The CPU's job is breadth; the GPU's is the specific job of drawing pixels. Both are essential, but they're different parts with different failure signatures.

Installing a CPU

1. Open the socket retention mechanism — lift the lever (LGA) or release the retention bracket (PGA), depending on the board.
2. Remove the socket cover (on LGA boards) and set it somewhere safe — you'll want it back if you remove the CPU later.
3. Align the CPU by its marker (a small triangle or a notch) with the matching mark on the socket, and lower it in straight and gently. It should sit flat with no force.
4. Secure the retention mechanism — lower the lever or bracket until it clicks. You should feel (and often hear) it lock.
5. Mount the cooler with fresh thermal paste (see the cooling section), and connect the cooler's fan to the CPU fan header on the board.

CPU cooling and seating

The CPU cooler — air or AIO, as covered in the cooling section — has to make firm, even contact with the top of the processor. The two classic seating failures: the cooler isn't fully locked down (uneven pressure, so part of the die isn't being cooled), or there's no thermal paste where there should be. Either one produces an immediately hot CPU, which can trigger thermal throttling (the CPU slowing itself down to stay safe) or a thermal shutdown. If a freshly built or repasted machine runs hot right away, the first thing to check is that the cooler is properly seated.

Integrated graphics: a diagnostic goldmine

Many CPUs include built-in ("integrated") graphics, which means the motherboard's own video output can display an image even with no separate graphics card installed. This is genuinely useful for diagnosis: if a system with a discrete GPU shows no display, and you can get an image on the motherboard's output using the CPU's integrated graphics, you've isolated the problem to the discrete GPU (or its seating/power) rather than the CPU, RAM, or the board. Not every CPU has this — some, like certain "F-series" Intel parts, don't — so check the specific CPU, but where it exists, it's one of the fastest ways to narrow down a no-display problem.

The graphics card renders images and video, handles gaming and GPU-accelerated work, and is very often the single biggest power draw and heat source in a system. That makes it the part most likely to expose a power supply's limits — and one of the most common sources of crashes and no-display problems.

What the GPU is for

The GPU takes the scene an application wants to show and turns it into the pixels on your screen. It's a separate card (a "discrete" GPU) rather than built into the CPU because dedicated silicon with its own memory is far better at the specialized job of rendering. The discrete card plugs into a PCIe slot on the board and, on most modern cards, needs one or more dedicated power connectors from the power supply on top of the power the slot provides.

VRAM

VRAM is the graphics card's own memory — separate from the system's RAM, and specific to the card. It holds the textures, frame buffers, and other data the GPU is actively working with. How much you need is workload-dependent: more is better for high-resolution textures and professional work, and modern games increasingly benefit from more VRAM. It's a spec to check when comparing cards, but the "right" amount depends on what you're doing and at what resolution — there's no single magic number.

Powering the GPU

A graphics card draws its power from two places: the PCIe slot (which supplies a limited amount) and one or more dedicated PCIe power connectors (6-pin or 6+8-pin) plugged in from the power supply. Two practical points: a card that needs two power connectors should ideally get two separate cables from the PSU rather than one cable split into two connectors, because a single cable has its own current limit. And a GPU that isn't getting enough power is a classic cause of crashes, artifacts, and shutdowns under load — which is why "check the GPU power connectors" is a standard step in diagnosing a graphics card that misbehaves. When you install a more powerful card than before, re-check that the power supply can actually feed it.

Installing a GPU

1. Power off and unplug, and remove any bracket screws where the card will sit.
2. Release the PCIe slot latch and seat the card firmly until the latch snaps back — it should be flush and secure.
3. Connect the PCIe power connector(s) from the power supply (separate cables where the card wants more than one).
4. Screw the card to the case bracket and close the panel.
5. Test — a big card can be awkward; having a second set of hands to support it while you connect the power is well worth it.

When the GPU is the suspect

The usual signs: visual artifacts (weird shapes, colors, or flickering on screen), crashes or shutdowns that happen specifically under graphical load, or no display on the GPU's own outputs. That last one is where the integrated-graphics test from the CPU section earns its keep — if you get an image on the motherboard's output but not the GPU's, the discrete card (or its seating or power) is the problem, not the rest of the system. If a monitor is involved in the no-display, the monitor no-signal guide covers sorting out whether it's the card, the cable, or the display. If GPU problems are surfacing as crashes, the blue screen guide helps read whether the stop code points at the GPU.

The case is both the container and the airframe. It holds everything in place and defines how air moves through the system. A good case with poor airflow is worse than a modest case with good airflow, so it's worth understanding what you're actually buying when you buy a case.

Form factors

Case size and board size have to match. The main board form factors are ATX (full size), Micro-ATX (smaller), and Mini-ITX (small). A case rated for a given form factor fits that board and any smaller one (an ATX case fits a Micro-ATX board), but not any larger one. The power supply also has a form factor (ATX PSU is the standard), and the case has to have room for it. When you're building, the case, the board, and the power supply all have to fit together — that's the first compatibility check.

Airflow design

The single biggest airflow factor is the front of the case. A mesh front breathes — cool air flows straight in to the intake fans. A tempered-glass front looks great but is a sealed wall that traps heat, so a glass-front case relies more on the top and rear for exhaust. Neither is wrong; they're a trade-off between looks and cooling. Whatever the case, the goal is the same as in the cooling section: a consistent intake-to-exhaust path, with the fans arranged to create it.

Cable management

Behind the board is a channel where the power and data cables are routed, out of the way of the airflow. Good cable management isn't just about looks — a tidy back of the board lets air flow where it needs to and makes the machine far easier to service later. Route the big cables (the 24-pin, the GPU power) along the channel, clip them where the case provides clips, and keep the front of the board as clear as you can. It takes a little extra time in the build and pays off every time you open the case again.

Choosing a case

Check four things before you commit: it fits your board form factor and power supply; it has the fan mounts you want (front, top, rear) for the airflow you're planning; it has room for your CPU cooler's height and your GPU's length (both are measured in the case spec); and its front is mesh or glass according to whether you value cooling or looks. A case that fits your parts and lets you build a clean airflow path is doing its job — the rest is preference.

A surprising share of "it doesn't work" after a build comes down to a cable that's in the wrong place or not fully seated. Cables are the cheapest part of the system and one of the most common failure points, so it's worth learning to tell them apart at a glance.

The main power cables

From the power supply, the cables that matter are the 24-pin ATX (the big flat connector to the motherboard), the CPU EPS (4- or 8-pin, labeled "CPU," to the top corner of the board), the PCIe power (6- or 8-pin, to the graphics card), and the SATA power (small, L-shaped, to the drives). The two that get confused are the CPU connector and the 24-pin — they're different plugs and not interchangeable, and the CPU one is the one people most often forget to plug in, which is why "check the CPU power connector" is a top "no POST" check.

Data cables

The SATA data cable is the thin, flat cable with L-shaped ends that connects a SATA drive to a SATA port on the board. It's easy to damage by bending the connector too far, so handle it gently. The front USB ports use small headers on the board (USB 2.0 and USB 3.0 are different connectors). Both are "connect and forget" once they're seated — but they do have to be seated.

How to tell them apart

The shape and the labels do the work. The 24-pin is by far the widest flat connector. The CPU EPS is a small 4- or 8-pin block. The PCIe power is a 6- or 8-pin block (a 6+8 is one connector with a separable 2-pin section). SATA power is a small, L-shaped 15-pin. When in doubt, the cable usually has a molded label, and the connectors have a keying that stops them going into the wrong socket — if it doesn't fit, it's the wrong one, so don't force it. The full "at a glance" list is in the reference tables below.

• Forgetting the CPU power connector — the machine is otherwise powered but the CPU has no power, so nothing happens.
• A 24-pin that's "in" but not clicked — it can sit partially engaged; make sure the latch locks.
• Splitting one PCIe cable to feed a card that wants two — use separate cables where the card needs more than one connector.
• A SATA data cable bent at the connector — the pins inside can break; if a port "stops working," inspect the cable's ends.

Almost every component swap follows the same workflow, and the part that saves the most trouble is testing in stages rather than building everything and then wondering what's wrong. Here's the general approach, and the principle behind it.

The general workflow

1. Power off and unplug the machine, flip the PSU switch off if it has one, and press the power button once to drain residual charge.
2. Ground yourself (an antistatic wrist strap, or touch the bare metal of the case before handling parts) and work on a clean, non-conductive surface.
3. Remove a side panel to get to the components, and keep your screws organized.
4. Work on one thing — install or replace a single component, fully seated and connected.
5. Test before closing up — power it on, confirm the change is recognized, and only then move to the next part.

The test-first principle

The most effective way to build or repair is to test as you go, so that if something fails you know exactly which step caused it. A common order: get the CPU, one stick of RAM, and the cooler in, and power on to see if the board posts (you may need integrated graphics or a GPU to see anything). Once that works, add the GPU, then the storage, testing after each. If it fails at a stage, you've isolated the problem to the part you just added, rather than facing a machine where any of six things could be at fault. It's slower in the moment and dramatically faster when something goes wrong.

First-boot checks

Before you press the power button on a fresh build or repair, run through the connectors: the 24-pin ATX is seated and latched, the CPU power connector is in, the RAM is clicked into its slot(s), the GPU is seated and its power is connected (if using one), and the front power-button header is on the correct pins. Nearly every "it doesn't turn on" or "no display" at first boot traces back to one of these. If it still doesn't post, the diagnosing section and the won't turn on guide walk through the next steps.

The most overlooked maintenance is also the most effective. A dusty PC runs hotter and louder, and heat — over months and years — degrades components and shortens their life. Ten minutes of cleaning every few months is the cheapest reliability upgrade there is.

How often to clean

It depends on the environment. A machine in a clean, air-conditioned office might be fine every six to twelve months; one on the ground floor of a house with pets may need attention every three to four months. The practical trigger is visual: when you can clearly see a coating of dust on the fan blades and inside the heatsinks, it's due. If a machine that used to run quietly and cool has gotten louder and warmer, dust is a leading candidate.

What you need

Compressed air (a can is fine), a soft brush (a clean, unused paintbrush or a watchmaker's brush), and, for repasting, isopropyl alcohol and lint-free wipes. That's most of it. You don't need water, and you don't need to disassemble the whole machine — the goal is to clear dust from the fans, heatsinks, and the power supply's intake filter.

How to clean safely

1. Power off, unplug, and open the case as for any other job.
2. Hold the air can upright and use short, controlled bursts — tipping it can spray liquid, which you don't want on the components.
3. Hold the fans still while you blow near them — a fan spun freely by compressed air can overspeed and damage itself (and act as a generator).
4. Clear the dust from the fan blades, the heatsink fins, and the mesh/filter, working from the inside out so you don't push dust deeper.
5. For the power supply's filter, clean the outside of the case's PSU intake vent only — never open the PSU itself.

What to avoid

• Water or liquid cleaners on the components — if it can conduct electricity, it doesn't belong in there.
• Spinning a fan free with compressed air — hold the blades.
• Opening the power supply to "clean the inside" — clean the external filter only; the inside is a hazard.
• Excessive force on cables, connectors, or fan blades — most damage in a clean is accidental and from pushing too hard.

The number one reason a part "doesn't work" is that it isn't compatible with what's already in the machine. Compatibility is checked before you buy, not after, and it's a small set of checks that saves a lot of disappointment.

The compatibility web

Each part has to fit the parts around it. The CPU has to match the motherboard's socket and chipset. The RAM has to be the right generation (and within what the board supports). The power supply has to have enough wattage and the right connectors for the graphics card. The case has to fit the board's form factor, the CPU cooler's height, and the GPU's length. And storage has to match an available SATA port or M.2 slot. None of these are hard to check — they just have to be checked, and they all connect to each other, which is why a build is a set of decisions made together, not a shopping list.

The three that always bite

Most compatibility failures come from one of three mistakes. Wrong RAM generation — buying DDR5 for a DDR4 board (they physically don't fit, but it's an easy order to get wrong). CPU and motherboard socket mismatch — a CPU for a different socket family than the board. And power supply — not enough wattage for the graphics card, or not having the PCIe power connectors it needs. If you check those three, you've cleared the vast majority of "it doesn't work" surprises.

How to check before you buy

For RAM, the motherboard manufacturer publishes a QVL (Qualified Vendor List) — the specific kits tested and confirmed to work on that board; it's the most reliable compatibility check there is. For a CPU, match the socket and chipset to the board. For a GPU, check its power requirements and its length against the power supply and the case. For a cooler, check the socket support and the maximum height the case allows. When in doubt, the manual and the manufacturer's own compatibility tools are the authoritative answers — "it should fit" is not the same as "it does."

Hardware diagnosis uses the same mindset as software diagnosis — isolate and test — but the tools are physical. The principle: change one variable at a time, use known-good parts, and let the process of elimination find the fault. Below are the common presentations and the order to work through them.

The diagnostic mindset

Two habits do most of the work. First, one variable at a time — if you swap three things and it starts working, you don't know which one fixed it, so you'll be guessing next time. Second, known-good parts — the fastest way to isolate a fault is to have a working RAM stick, a working GPU, and a working power supply to swap in. If a part works in another machine but not this one, the problem is here; if it fails in both, the part is the problem. That single test — "does it work elsewhere?" — resolves more hardware questions than anything else.

PC won't power on

Work from the wall inward. Is the outlet live (try another device)? Is the cable good? Is the power supply's rear switch on? Are the 24-pin and CPU power connectors fully seated? Is the front power-button header on the correct pins? If the whole power chain is correct and still nothing happens when you press the button, the power supply is the prime suspect — confirm with a known-good unit. A "no power at all" machine is usually one of these, in roughly this order, and it's almost never the CPU or the RAM. The won't turn on guide covers this in depth.

Powers on, no display

The most common hardware fault, and it's almost always RAM first, then the GPU. Reseat the RAM — try one stick in a known-good slot. Then reseat the GPU and confirm its power connectors are in. If the CPU has integrated graphics, temporarily plug a monitor into the motherboard's output: if you get an image there, the discrete GPU (or its seating or power) is the problem and the rest of the system is fine. If you still get nothing, try a CMOS reset (clearing the BIOS via the jumper or by removing the coin battery for a minute), and check that the CPU power connector is seated. Beep codes or LED indicators on the board, where present, can point you straight at RAM or the CPU.

Overheating

Overheating is a cooling-path problem, and the path has a few links: the fans (are they all spinning?), the dust (is the cooling clogged?), the paste (is it old or the cooler loose?), and the airflow (is air actually moving through the case?). Check the actual temperatures first (the cooling section has the ranges), confirm every fan is turning, clean out dust, and verify the CPU cooler is firmly seated. If temperatures were fine and then jumped, suspect the thermal paste or a fan bearing. A machine that shuts down after running a while and getting hot is thermally protecting itself — the fix is cooling, not a new part.

Random shutdowns and restarts

Two big suspects, and the clue is timing. If the machine shuts down or reboots immediately or under heavy load, it points at the power supply — it can't sustain the draw. If it shuts down after running a while and getting hot, it points at cooling — a thermal shutdown. The way to tell them apart: check the temperatures under load (if it's hot, it's heat; if it's cool and still dies under load, it's power), reseat the power connectors, and if it's power-related, a known-good PSU is the test. The keeps restarting guide covers the software side to rule that out first.

Storage problems

If a drive has disappeared or slowed down, start with the connection — the SATA data and power cables, or the M.2 seating. A clicking hard drive is a classic mechanical-failure sign: back up immediately and plan to replace it. For an SSD, check its reported health in the operating system. A useful test: if a drive works in another machine but not here, the problem is the port or the cable, not the drive; if it fails in both, the drive is dying. The hard drive not detected guide walks through isolating these.

Memory problems

RAM problems show as a machine that won't boot, beep codes, or random BSODs and instability. The method is the same every time: test one stick at a time, in a known-good slot. If you enabled XMP/EXPO and it's unstable, disable it and test at the base speed first. A stick that works in one machine but not another is usually the board or the slot. RAM is one of the cheapest and fastest things to rule out, which is why it's so high on the "won't boot" list. If memory issues are surfacing as crashes, the blue screen guide helps read whether the stop code points at memory.

If you have a specific problem in front of you, pick the closest match and jump to the right checks. If you'd rather not use the tool, the tables and sections above cover every path — nothing here is only available through the selector.

The best upgrades are planned ones. A part that fits, is compatible, and is matched to what you actually want to do is a good upgrade; the same part bought without checking can be a pile of money that doesn't work. Here's how to think about it.

The highest-impact upgrades

For a typical system, the two upgrades with the biggest "everything feels better" effect are storage (moving the operating system from an HDD to an SSD) and memory (enough RAM so the system isn't swapping to disk). After those, the next upgrade depends entirely on what you do: gaming points at the graphics card, and general longevity points at cleaning and repasting. A common mistake is spending on the part that looks most impressive rather than the one that's actually the bottleneck for your use.

Upgrade by goal

If the goal is gaming or GPU work, the graphics card is the part — but it's also the upgrade most likely to outgrow your power supply, so check the wattage and connectors before you buy. If the goal is general speed, an SSD for the boot drive plus enough RAM is the combination. If the goal is longevity and quiet, a clean, a fresh thermal paste, and a well-arranged set of case fans will do more than any single new part. Matching the upgrade to the goal is what makes it worth the money.

The compatibility check, again

Before you order, run the compatibility checklist in the reference tables. The GPU is the one to be extra careful with: it has to fit the case (length), fit the board (a free PCIe x16 slot), and be fed by a power supply with enough wattage and the right PCIe connectors. If the power supply is the limiting factor, the "upgrade" is really a two-part purchase (the new card and a new power supply), which is worth knowing before you commit to the card.

Plan the power

Whenever an upgrade increases the system's power draw — and a new graphics card is the classic case — re-do the wattage math from the power section. An upgrade that quietly pushes the system past its power supply's comfortable range is a recipe for the exact "random shutdowns under load" problem described in the diagnosing section. Check the power first, and you avoid buying a part you can't actually run.

Two enemies of the parts in your hands are static electricity and live power. Both are easy to protect against with a few habits, and both are the reason a part that looked fine can be killed by the handling without any visible sign of damage.

Static (ESD)

Static discharge can damage or destroy a component — and often does so without any visible mark, so the part looks fine and simply fails later or intermittently. You protect against it by giving the static a safe path to ground: an antistatic wrist strap clipped to the bare metal of the case is the proper tool, and simply touching the bare metal of the case before you handle parts is a reasonable substitute. Work on a clean, hard, non-conductive surface (a wooden or laminate desk, not a carpet), and avoid handling components by their pins or gold contacts — hold them by the edges.

Power safety

Power the machine off, unplug it, and — for desktops — flip the power supply's rear switch to off. Then press the front power button once: this drains the residual charge held in the capacitors so there's no surprise voltage when you start moving things. Don't work on a machine that's connected to power, and don't open the power supply itself (it holds a dangerous charge and mains-voltage circuitry, even when unplugged). For laptops, remove the battery where it's removable, and treat a swollen battery as a hazard to be handled professionally, not something to pry at.

Handling components

Hold parts by the edges, not the contacts. Don't touch the CPU pins (on the board or the chip) or the gold fingers on RAM and cards. Support heavy parts (a large graphics card) so their weight isn't on the connector. And never force a connector — the keying exists to stop it going in the wrong way, so if it's resisting, it's misaligned, and forcing it is how pins and ports get broken. When in doubt about how much force is right, the answer is "less than you think."

The two tables worth having in front of you while you work: the cables at a glance, and the compatibility checklist to run before you buy. The component, storage, cooling and troubleshooting tables are in their sections above.

Cables and connectors at a glance

Compatibility checklist (run before you buy)

Useful commands to see what's installed

Before or after a hardware change, a couple of built-in tools let you confirm what the system actually sees — the CPU and RAM, the graphics card, and the health of the drives. These are read-only and safe to run.