A loud computer is not necessarily an overheating computer. Before you buy thermal paste or start disassembling anything, get an actual temperature reading — it takes about three minutes and it tells you whether you have a heat problem, a software problem, or just a fan curve someone set aggressively.
Download Core Temp or HWMonitor, both free. Core Temp is lighter and shows CPU core temperatures directly. HWMonitor shows CPU, GPU, drives, and fan speeds in one window, which is more useful for diagnosis. Install it, open it, and watch two numbers: the temperature at idle with nothing running, and the temperature after ten minutes of whatever work makes the machine misbehave.
Watch out for one gotcha during install — HWMonitor’s installer has occasionally bundled optional extras. Uncheck anything that isn’t the program itself.
What the numbers should look like
| Component & state | Normal | Watch it | Act now |
|---|---|---|---|
| Desktop CPU, idle | 30–45 °C | 50–60 °C | above 65 °C |
| Desktop CPU, sustained load | 60–80 °C | 80–90 °C | above 95 °C |
| Laptop CPU, idle | 40–55 °C | 55–65 °C | above 70 °C |
| Laptop CPU, sustained load | 70–90 °C | 90–95 °C | pinned at 99–100 °C |
| GPU, idle | 30–45 °C | 50–60 °C | above 65 °C |
| GPU, sustained load | 65–83 °C | 83–90 °C | above 90 °C |
| Hard drive / SSD | 25–45 °C | 45–55 °C | above 60 °C |
Two things to keep in perspective. Modern CPUs are designed to hit 85-95 °C under heavy load and that’s within spec — brief spikes when you launch something are normal and harmless. Laptops legitimately run hotter than desktops because there’s nowhere for the heat to go. A laptop at 88 °C while exporting video isn’t broken.
What’s not fine is a machine sitting at 70 °C doing nothing, or one that hits 100 °C and stays there. Around 100 °C most processors begin thermal throttling — deliberately cutting clock speed to protect themselves. That’s the mechanism behind the most telling symptom of all.
How to tell it’s heat and not something else
The signature symptom is performance that’s fine for the first few minutes and then falls off a cliff. Boot up, everything’s snappy. Ten minutes into real work, it’s crawling. Close everything, let it sit, and it’s fine again. That pattern is thermal throttling and almost nothing else produces it.
Other reliable signs:
- Fans at maximum constantly, including when the machine is idle at the desktop.
- Abrupt shutdowns with no warning — no error, no blue screen, just off. That’s the thermal cutoff firing. It’s a protective feature, and it means temperatures went past where throttling could save it.
- The chassis is genuinely hot, not warm. If you can’t comfortably keep a hand on the underside of a laptop or the top of a desktop case, that’s real.
- Exhaust air is weak even though the fan sounds loud. Points to a blocked heatsink.
- Blue screens under load, especially WHEA_UNCORRECTABLE_ERROR or CLOCK_WATCHDOG_TIMEOUT. Heat isn’t the only cause but it’s a common one.
- Graphical artifacts — flickering, colored blocks, texture corruption — under load usually points at GPU temperature specifically.
Now the counter-check. Before assuming heat, open Task Manager with Ctrl + Shift + Esc, click the Performance tab, and watch CPU utilization. If you’re sitting at 90% CPU with no apps open, you don’t have a cooling problem — you have a software problem that’s generating the heat. Fixing the cooling on a machine whose real issue is a runaway process just makes it a quieter machine that’s still slow.
Causes, roughly in order of likelihood
Dust
Far and away the most common cause, and the one people underestimate. Dust doesn’t just coat things — it packs into a felt-like mat between the fan and the heatsink fins, and once that mat forms, airflow drops close to zero while the fan still spins and sounds busy.
Two to three years in a normal office does it. A carpeted room, a machine on the floor, a workshop, a pet in the building, or anyone who vapes indoors accelerates it dramatically. Vape residue is particularly bad — it’s sticky, and it glues dust into place.
Cleaning a desktop properly
Shut down, unplug the power cable, and move the machine outside or to a garage. Do not do this at a desk — you will simply relocate the dust into the room and breathe a fair amount of it.
Open the side panel. Get canned air, and hold each fan’s blades still with a finger or a zip tie before you blow air through it. This matters: spinning a fan with compressed air drives it far past its rated speed and can generate enough voltage through the motor to damage it or the board it’s connected to. Then work through the CPU heatsink fins, the GPU shroud, the power supply intake, the case filters, and the front intake vents.
Keep the can upright and use short bursts — tipping it sprays freezing liquid propellant onto components. Hold the nozzle a few inches back.
Do not use a household vacuum inside a computer.
Plastic vacuum nozzles build static charge against moving air and dust, and a discharge onto a motherboard can kill components outright — often not immediately, which makes it hard to connect cause and effect later. Use canned air or a proper anti-static electric duster. A vacuum held near the case to catch dust as it leaves is fine; inside the case it is not.
Cleaning a laptop
Blowing air into a closed laptop’s vents mostly pushes the dust mat deeper into the heatsink. It buys a little time and sometimes makes things worse. Proper laptop cleaning means opening the bottom panel, removing the fan, and physically pulling the packed dust off the radiator fins.
If you’re comfortable with small screws and ribbon cables, that’s a 30-45 minute job. If not, it’s worth having done — it’s routine work and it’s the single most effective fix for a laptop that’s been throttling for a year.
Blocked airflow
Free to fix and surprisingly often the whole problem. Laptops pull intake air from the underside. Set one on a bed, a couch cushion, a blanket, or even a thick desk mat and you’ve sealed the intake. Temperatures can climb 15-20 °C from that alone.
Use a hard flat surface, or a stand that lifts the rear. A cheap laptop cooling pad with fans is genuinely worth the $25 for a machine that runs hot.
For desktops: keep at least four inches of clearance behind the exhaust and at the front intake. A tower crammed into a closed cabinet under a desk recirculates its own hot air. Off the carpet, too — carpet fibers get sucked straight into the bottom-mounted PSU intake.
Dried thermal paste
Thermal paste fills microscopic gaps between the processor lid and the heatsink base. It dries out and loses conductivity over roughly three to five years — sooner on a machine that runs hot daily, sooner still on the cheap grey paste used in budget builds.
The tell: the fan is clean, the vents are clear, the machine still overheats. Especially if it’s five-plus years old and has never had this done.
Repasting means removing the heatsink, cleaning both mating surfaces with 90%+ isopropyl alcohol and a lint-free cloth, applying a pea-sized dot of new compound, and remounting with even pressure. Materials cost under $15. A good repaste commonly drops load temperatures 10-20 °C, which is a bigger improvement than most hardware upgrades deliver.
It’s also the point where a lot of people should hand it off. Over-tightening a heatsink or knocking a surface-mount component off a board turns a $15 job into a new machine.
A fan that’s failing or dead
Bearings wear out. A dying fan announces itself with a grinding or clicking noise, a rhythmic ticking, or a whine that changes pitch — or with silence where there should be noise.
HWMonitor reports fan RPM. A CPU fan reading 0 RPM while temperatures climb is a dead fan or a disconnected header. Look at the machine while it’s under load and confirm each fan is actually turning. Case fans are $10-20 and swap in minutes; CPU cooler and laptop fans are model-specific but usually available.
On an AIO liquid cooler, also listen for a gurgling or rattling pump. A failed pump means the fans spin, the radiator stays cool, and the CPU cooks — a confusing failure because everything looks like it’s working.
The room
Cooling can only move heat into the surrounding air. Every degree the room gains, the machine gains. A closet-turned-server-room, a west-facing office in August, a desk against a sunny window — these are real causes, not excuses.
If you have a server or NAS in a small unventilated room, take it seriously. That equipment runs continuously and generates continuous heat, and a closed closet will reach 100 °F ambient without much trouble. Ventilate it or move it.
Software burning CPU
Task Manager’s Performance tab tells you in seconds. Common offenders: a browser with 60 tabs and a runaway ad script, Windows Search reindexing after an update, Windows Update itself, two antivirus products fighting each other, a stuck print spooler, or OneDrive/Dropbox re-syncing a large folder.
Most of these settle within an hour. If something has been at 80-100% CPU for days, that’s not normal maintenance.
Cryptominers and malware
Worth calling out separately because it’s tailored to be hard to notice. Cryptojacking malware uses your CPU or GPU to mine currency, and better-written versions throttle themselves down when you move the mouse and ramp back up when you walk away.
The pattern to look for: the machine is hot and the fans are screaming when you come back to it, then it settles once you start using it. Or a process you don’t recognize consuming heavy CPU. Or high usage that shows up in Resource Monitor but not obviously in Task Manager.
Run a full scan with Windows Defender plus a second opinion from Malwarebytes. Browser-based miners run inside a tab — if closing the browser stops it, that’s your answer, and an ad blocker plus a look at installed extensions will finish the job.
Power supply problems
Less common, but it shows up in machines where someone added a graphics card to a prebuilt system without upgrading the PSU. An undersized or aging power supply runs hot, dumps that heat into the case, and can cause shutdowns under load that look exactly like thermal shutdowns.
Clue: the shutdown happens the instant load spikes, temperatures were reasonable a second before, and the machine sometimes won’t power back on immediately. A smell of hot electronics near the rear vent is a stop-using-it-now signal.
Do this in order
- Install Core Temp or HWMonitor. Record idle and load temperatures. Now you have a baseline to measure against.
- Check Task Manager. If something is eating the CPU, deal with that first.
- Clear physical obstructions — get it off the carpet, off the cushion, out of the cabinet, four inches of clearance.
- Run a full malware scan.
- Clean the dust out, properly, outdoors, holding the fan blades still.
- Re-measure. If temperatures dropped meaningfully, you’re done.
- Still hot and the machine is 3+ years old? Repaste.
- Still hot after that? Check every fan’s RPM and replace what isn’t spinning correctly.
When to stop and get it serviced
Some of this stops being a DIY job:
- The machine shuts down within a couple of minutes of booting — you don’t have time to work on it, and continuing to run it risks the CPU.
- You smell burning or see any browning or bulging on the board or capacitors. Unplug it and stop.
- It still throttles after cleaning and repasting — likely a warped heatsink, a bad mount, or a cooler that was never adequate for the chip.
- It’s a business machine with your only copy of working data on it. Heat kills drives, and a failing drive plus no backup is a much bigger problem than a noisy fan.
- It’s a laptop still in warranty. Opening it may void coverage. Call the manufacturer first.
For a business, the practical framing is that thermal throttling is a productivity tax you’re paying quietly. A workstation running at 60% of its speed all afternoon costs more in wasted time over a year than the cleaning would. Building an annual clean-and-check into your maintenance routine — the same way you’d change filters on the HVAC — is cheap insurance.
Common questions
Is 90 °C too hot for a CPU?
Depends on what it’s doing. Ninety degrees during a video export or a long compile is within spec for most modern processors and won’t damage anything. Ninety degrees while you check email is a serious problem. The number alone means little — the number paired with the workload is what tells you something. Sustained 100 °C with visible throttling is where you should stop and fix it.
Can overheating permanently damage my computer?
Modern CPUs and GPUs protect themselves by throttling and then shutting down, so outright cooking a processor is rare now. The real damage is cumulative and lands elsewhere — repeated heat cycling degrades capacitors and solder joints over time, and mechanical hard drives are genuinely sensitive to sustained heat, with failure rates climbing noticeably above 50 °C. Years of running hot shortens the machine’s life even if nothing dramatic ever happens.
How often should thermal paste be replaced?
Every three to five years for a typical office machine, closer to three for one that runs demanding work daily. Quality paste in a cool, low-load environment can last longer. Don’t repaste on a schedule for its own sake — do it when temperatures have climbed and cleaning didn’t bring them back down. That’s the actual trigger.
Do cooling pads for laptops actually help?
Modestly. Expect somewhere between 3 and 8 °C on a machine with clean, unobstructed vents — worthwhile but not transformative. The larger benefit is that the pad elevates the laptop and guarantees the underside intake isn’t sealed against a surface. If the internal heatsink is clogged with dust, no pad will compensate; clean it first, then add the pad.
My fans are loud but the temperatures look fine. What’s going on?
Usually a fan curve set too aggressively, which you can adjust in UEFI setup — look for Smart Fan, Q-Fan, or Fan Control and set a gentler profile. Sometimes it’s a bearing starting to fail, which changes the sound quality rather than the volume. And sometimes it’s one specific fan mounted where it resonates against the case. If the temperatures are genuinely good, this is a noise problem, not a heat problem, and it’s fixable without touching the cooling.