Reference
Safe GPU and CPU temperatures on Linux
A plain reference for what your sensors are actually telling you, reading from Linux hwmon and NVML, the same sources lm-sensors, MangoHud, and Thermal Canary Pro all read from. These are general guidance for consumer hardware, not a substitute for your specific card or CPU's documented limits, cooler design and case airflow vary enough that your safe range can shift a few degrees either way.
GPU temperatures
| Range | What it means |
|---|---|
| 30 to 50°C | Idle or light desktop use, normal |
| 60 to 80°C | Typical range under sustained gaming load, normal |
| 80°C average, sustained | Running consistently warm, worth checking case airflow or fan curve |
| 88°C or above, sustained | Critical: thermal throttling is likely already active |
Hotspot or junction temperature reads 10 to 15°C above the main GPU temperature under full load on plenty of consumer cards, that gap alone is not a fault. What matters is whether the gap grows over weeks: see the guide onGPU hotspot delta and when to repaste.
CPU temperatures
| Range | What it means |
|---|---|
| 30 to 45°C | Idle or light desktop use, normal |
| 60 to 80°C | Typical range under sustained gaming or all-core load, normal |
| 85°C average, sustained | Running consistently warm, worth checking cooler contact or fan curve |
| 92°C or above, sustained, or active throttling | Critical |
Recent Intel and AMD Ryzen parts are designed to run close to their thermal junction maximum for short bursts under heavy all-core load, a brief peak during a benchmark is expected behavior. A sustained average near the critical range, across a whole gaming session, is the more useful thing to watch.
Why "sustained" matters more than a single peak
One high reading during a load spike tells you almost nothing on its own, load spikes are normal and coolers are designed to absorb them. A sustained average across a session tells you something real about your cooling, and a peak that crosses the critical line and stays there is what actually correlates with throttling and long-term wear. That is the distinction behind every threshold in this reference: the warm-state warnings above trigger on an averaged reading, the critical thresholds trigger on a sustained peak.
Where these numbers come from
These are the exact thresholds Thermal Canary Pro's own Health Diagnostic and Smart Alert Overlay use internally, documented here so they are useful whether or not you run the app. If you want them tracked automatically, with a silent toast only when a threshold is actually crossed, and a 60-second sensor history auto-saved if your GPU crashes, that is what Pro adds on top of raw sensor readouts.
FAQ
What GPU temperature is dangerous?
As a general rule for consumer GPUs, a sustained edge temperature at or above 88°C is worth treating as critical: it usually means thermal throttling is already active and the card is being protected by its own firmware. A brief spike into the high 80s under a stress test is not automatically a problem, sustained is what matters, not one reading.
Is 80°C too hot for a GPU under load?
No, 80°C is normal for a lot of consumer GPUs under sustained gaming load, especially smaller-form-factor cards. It is worth watching if it becomes your average rather than your peak, since that points at reduced airflow or a fan curve that is too conservative, but it is not itself a critical temperature.
What CPU temperature is too high?
Sustained readings at or above 92°C, or any point where the CPU is actively thermal throttling, count as critical. Modern CPUs (recent Intel and AMD Ryzen parts) are designed to run close to their thermal junction maximum under heavy all-core load for short periods, so a brief peak near that range during a benchmark is expected behavior, not damage in progress.
Why does GPU hotspot or junction temperature read higher than the main GPU temperature?
Because it is measuring a different point on the die. The primary GPU temperature (often called edge temperature) is an average across the whole chip, while hotspot or junction temperature is the single hottest point. A gap of 10 to 15°C between the two is routine at 100% load on plenty of consumer cards. What actually matters is whether that gap grows over weeks, since a widening delta is the classic sign of drying thermal paste.
Does a high peak temperature matter as much as a high average temperature?
Less than you would think. A single high peak during a heavy load spike is normal thermal behavior. A high average, sustained across a whole session, is the more useful signal that airflow, fan curve, or cooler contact needs attention, which is why sustained-average thresholds are used for early warm-state warnings and peak thresholds are reserved for critical alerts.
General guidance for consumer desktop hardware. Always defer to your GPU or CPU manufacturer's own documented thermal limits where they are published, cooler design and case airflow both shift what counts as normal for your specific setup.