Windows · Guide
How to Monitor CPU and GPU Temperatures on Windows
Short answer
Install HWMonitor or HWiNFO and read the CPU package and GPU sensors while the PC is under load. Anything sustained under roughly 85 °C on a laptop CPU and 80 °C on a desktop GPU is normal; throttling and shutdowns come from sustained higher readings.
Requirements
- Windows 11 or Windows 10
- HWMonitor or HWiNFO (free)
- Optional: GPU-Z for graphics-card detail
- A repeatable load, such as a game or an encode
Steps
1. Record an idle baseline first
Open HWMonitor with nothing else running and note the CPU package and GPU temperatures after two minutes. Idle readings well above 50 °C already point to dust or dried thermal paste.
2. Identify the right sensors
Use CPU Package (or Tctl/Tdie on AMD) rather than individual core values, and the GPU Temperature or Hot Spot value from your graphics card section. Motherboards expose several unrelated zones.
3. Apply a realistic load
Play the game or run the workload that actually causes the problem for at least 15 minutes. Synthetic stress tests push harder than real use and can make a healthy PC look faulty.
4. Watch the maximum column, not the instant value
HWMonitor keeps a Min and Max per sensor. The Max after a full session is the number that matters, since brief spikes are normal.
5. Log a longer session with HWiNFO
In HWiNFO sensors view, start logging to CSV before a long session. Reviewing the log afterwards shows exactly when clocks dropped and which sensor rose first.
6. Correlate temperature with clock speed
Thermal throttling looks like a temperature ceiling with falling clocks. If clocks drop while temperatures stay moderate, the limit is power or the fan curve rather than heat.
7. Fix the cause, then re-measure
Clear dust from intake and heatsink fins, raise a laptop for airflow, check fan curves, and repaste an older machine. Repeat the same workload afterwards to confirm the change worked.
Alternative methods
1. Use Task Manager for GPU basics
The Performance tab in Task Manager on Windows 11 shows GPU temperature for many discrete cards, which is enough for a quick sanity check.
2. Use Speccy for a readable snapshot
Speccy lists temperatures alongside full hardware detail, which is handy when you want one screenshot to send to someone else.
Troubleshooting
One sensor reads a wildly impossible value
Unlabelled motherboard zones report nonsense on some boards. Trust the CPU package and GPU sensors and ignore generic TMPIN entries that never change with load.
No GPU temperature appears at all
Laptops with hybrid graphics only report the discrete GPU while it is active. Start a 3D workload, or read the card directly in GPU-Z.
Temperatures are fine but the PC still shuts down
Sudden shutdowns without heat usually mean a power supply or memory fault. Log voltages in HWiNFO and check Event Viewer for kernel-power entries.
Frequently asked questions
What CPU temperature is too high?
Modern CPUs are rated to run near 95-100 °C before throttling, but sustained operation at that ceiling means cooling cannot keep up. Aim for headroom under load rather than a single number.
Is HWMonitor or HWiNFO better?
HWMonitor is quicker to read for a one-off check. HWiNFO exposes far more sensors and can log to CSV, which is what you want for diagnosing intermittent throttling.
Do monitoring tools slow the PC down?
Polling sensors costs almost nothing, but very short polling intervals plus on-screen overlays can add measurable overhead in games. One second is plenty.
Can I monitor temperatures in a virtual machine?
No. Hardware sensors are not exposed to guests, so temperature readings must be taken on the host operating system.