How to check PC temperatures - free tools and what numbers matter

Most PC problems that get blamed on software are actually thermal. A CPU that throttles under load, a GPU that crashes mid-game, a system that shuts off without warning - temperature is the first thing to rule out, and it takes five minutes with a free tool.

The tools worth using

HWiNFO64 is the most comprehensive free option. It shows temperatures, voltages, fan speeds, and power draw for every sensor in your system, updated in real time. Download it from hwinfo.com and run it in Sensors-only mode.

MSI Afterburner is the go-to for GPU monitoring specifically. It can display temperatures, clock speeds, and usage as an overlay inside games, which is useful for catching problems that only appear under load.

Core Temp is a simpler option focused purely on CPU temperatures, good if you just want a quick reading without the full HWiNFO dashboard.

All three are free, don't require installation of anything else, and won't interfere with each other if you run more than one.

What numbers to look at

There are a lot of sensors in HWiNFO. The ones that actually matter:

CPU temperature - look for "CPU Package" or "Tdie" depending on your platform. This is the overall CPU temperature, which is more useful than individual core readings for general health checks.

GPU temperature - labelled "GPU Temperature" in most tools. Straightforward.

GPU hot spot / junction temperature - a separate sensor on AMD cards and some Nvidia ones. This measures the hottest point on the GPU die rather than the average. It runs higher than the main GPU temperature and that's normal - it's the number that matters for actual thermal safety.

VRM temperatures - found in HWiNFO under motherboard or CPU power sensors. Often overlooked, but a VRM running hot is a real problem that won't show up in CPU temperature until it's too late.

NVMe SSD temperature - SSDs throttle when hot, which causes stuttering that looks like a CPU or RAM issue. Worth a glance if you're troubleshooting slowdowns.

What's normal, what's a concern, what's a problem

CPU under load:

• Below 80°C - fine
• 80-90°C - warm, worth improving cooling or repasting
• Above 90°C - throttling is likely, needs attention
• Above 95°C - thermal shutdown territory on most chips

CPU at idle:

• Should be 10-20°C above ambient room temperature. A CPU idling at 60°C+ in a cool room points to a cooling or paste problem.

GPU under load:

• Below 83°C - fine for most cards
• 83-90°C - acceptable but on the warm side
• Above 90°C - fan curve or airflow issue, worth investigating
• Above 95°C - throttling likely, check fans and case airflow

GPU hot spot / junction:

• AMD cards: up to 110°C is within spec, though lower is better
• Nvidia cards: junction data varies by model - above 95°C average is worth monitoring

VRM temperatures:

• Below 90°C - normal
• Above 100°C - concerning, especially on budget boards under sustained CPU load
• Above 110°C - a real problem that will cause instability and accelerated wear

NVMe SSD:

• Below 60°C under load - fine
• Above 70°C consistently - throttling likely, consider an M.2 heatsink

How to test under load, not just at idle

Idle temperatures tell you very little. The readings that matter are under load, because that's when thermal problems appear.

For CPU: run Prime95 (small FFTs preset) or Cinebench R23 for 10-15 minutes while watching HWiNFO. Both push the CPU hard and will expose any thermal headroom issues quickly.

For GPU: run FurMark or play a demanding game for 20-30 minutes with MSI Afterburner's overlay active. Watch for temperatures climbing steadily rather than levelling off - steady climbing with no plateau suggests the cooling can't keep up.

What to do if temperatures are too high

Clean the PC first. Dust in heatsink fins is the single most common cause of high temperatures and the easiest fix. See our guide on how to safely clean a PC.

Check fan operation. Fans should ramp up as temperatures rise. If a fan isn't spinning or is spinning slowly at high temperatures, check the fan curve in BIOS or whether the fan itself has failed.

Replace thermal paste. Thermal paste dries out and loses effectiveness over time, typically after 3-5 years. A CPU that ran cool when new but runs hot now is often just in need of a repaste - it's a 20-minute job that can drop temperatures by 15-20°C on an older system.

Improve case airflow. A case with poor intake-to-exhaust balance traps hot air. Make sure you have more intake than exhaust fans, and that there's a clear path for air to flow from front to back and bottom to top.

When to send it in

If temperatures are high after cleaning and you're not comfortable removing heatsinks to repaste, that's a straightforward job we can do on the bench. If temperatures are fine but the system is still crashing or throttling, the problem is likely elsewhere - component-level diagnosis can identify whether it's a power delivery issue, a failing VRM, or something more specific that temperature monitoring alone won't catch.