CPU and GPU temperatures while gaming: what is normal

Your processor sits at 95 °C in a game, or a graphics card's hotspot reads past 100 °C, and the number looks like a warning. Whether it is one depends on two things: the limit the manufacturer publishes for that exact chip, and whether the chip is pulling its clocks down to stay under it. This page covers how to read both, what the vendors publish, and the physical causes to check when a PC really does run too hot.

How to read the temperatures

Read them while the game is running, under the load that worries you.

• GPU: Task Manager. For a dedicated graphics card, Task Manager > Performance shows the current temperature next to the GPU's listing. When Microsoft added it, its notes said it needs a driver supporting WDDM 2.4 or later and shows Celsius only. To see your driver model, press Win+R, run dxdiag and read Driver Model on the Display tab.
• CPU: HWiNFO. A free tool for real-time monitoring of CPUs, GPUs and mainboards, with logging. A log taken through a gaming session shows the peak and when it happened, not just the reading after you have already left the game.

w0a's Home page also shows live CPU and GPU clocks and temperatures, and it is part of the free tier (what is free); its CPU temperature reading comes from an optional third-party module that asks you separately (why it asks).

What Intel and AMD publish for a CPU

Intel defines Tjunction max as "the maximum thermal junction temperature that a processor will allow prior to using internal thermal control mechanisms to reduce power and limit temperature", and says the limit "varies per product and usually is between 100°C-110°C" (Intel). Intel's spec pages list a maximum operating temperature of 100 °C for the Core i9-14900K and 105 °C for the Core Ultra 9 285K. Intel is also direct about what reaching it means: "Being at maximum temperature while running a workload isn't necessarily cause for concern."

AMD lists a Max. Operating Temperature (Tjmax) on its product pages: 95 °C for the Ryzen 9 7950X (AMD), the Ryzen 9 9950X and the Ryzen 7 9800X3D, but 89 °C for the Ryzen 7 7800X3D. Take the figure for your exact model from its product page, not from a rule of thumb for the family.

What the GPU makers publish

NVIDIA lists a Maximum GPU Temperature for each card: 90 °C for the GeForce RTX 5090 and RTX 4090, 88 °C for the RTX 5070 Ti and 85 °C for the RTX 5070 (NVIDIA). These figures describe NVIDIA's Founders Edition or reference design, and NVIDIA notes that specifications "may vary by add-in-card manufacturer", so check your card maker's page too.

Some cards report a second temperature. Testing a Radeon RX 5700 XT, TechPowerUp noted that the GPU "reports a second temperature called 'Hot Spot' or 'Junction Temperature'", and measured the hotspot at 110 °C while the GPU temperature was 92 °C (TechPowerUp). That is one reviewer's measurement of one card, not a published figure for any Radeon. What it does show is that the two numbers are not interchangeable: in that test the hotspot ran 18 °C above the GPU temperature, so do not hold a hotspot reading against a limit published for the GPU temperature.

What throttling is, and how it shows

Throttling is the protection working, not the failure. Dell's definition: "When the processor or graphics chip reaches a high temperature, the computer reduces clock speed to lower heat output", and "Throttling by itself does not indicate a fault" (Dell). Intel says the processor "typically reduces frequency and power to prevent overheating", and if throttling cannot hold a safe temperature it "will automatically shut down to prevent permanent damage".

Intel's list of overheating symptoms is the practical version:

• the reported CPU frequency is lower than expected, or there is evidence of throttling
• the whole system feels slow
• fan noise is excessive
• the system boots, then shuts down automatically after a short time

Dell draws the line between normal and not: throttling "that starts within minutes of light use, or that occurs together with lock ups or shutdowns, indicates a problem".

Dell adds that during "sustained heavy work such as gaming" a computer "can become slower even when everything is working as designed". In a game, that slowdown is a frame-rate drop, and Sudden FPS drops in games covers telling heat apart from power-state changes, driver resets and background work. If the PC switches off or restarts instead, start with PC restarts or shuts down while gaming.

What to check

Intel writes its checklist for self-built PCs; for a PC bought complete, laptops included, it recommends contacting the manufacturer.

1. Dust. Intel's checklist asks you to look for dust build-up on all fans and to keep intake and exhaust vents clear. Dell explains why: dust on the cooling fins "prevents air from flowing around the fins, so the fan works harder and louder".
2. Fans. The cooler's fan powered, its impeller free, its inlet not blocked by components or cables (Intel). A fan that does not speed up is not necessarily broken: some Dell laptops take a thermal profile from the BIOS as well as from Windows software, and Dell notes that with the BIOS set to a quiet or performance-oriented mode "the fan might not respond the way the software profile suggests". Loading default BIOS settings is on Intel's list.
3. Case airflow. Intel asks for adequate airflow inside the case, cables that do not obstruct it, and a balance between intake and exhaust.
4. Paste and mounting. Intel's checklist: the protective layer removed from the cooler's thermal material, the right amount of thermal interface material, all anchors fastened evenly. Dell lists a failing fan, heat sink or thermal material among the hardware causes.
5. A laptop on a soft surface. Dell: "A hard, flat, non-cloth surface allows air to reach the fan intake. Using a Dell laptop on a blanket, pillow, cloth, or paper can cover the fan air intake and cause overheating." A cooling pad's own fans "supplement the fans in the laptop".
6. The room. Dell classes a computer that is warm because it is used "in direct sunlight or a hot room" as usage-related, not a fault, and gives 0 °C to 35 °C as the recommended room temperature for its laptops, with exact limits in each model's guide.

What does not help

• Chasing a lower number on a chip that is inside its published limit and not throttling. Intel's sentence bears repeating: being at maximum under load "isn't necessarily cause for concern".
• Judging by one reading taken after the game has closed. Log during play.
• Expecting a setting to fix a physical cause. No software removes dust from a heatsink, reseats a cooler or lifts a laptop off a blanket; it can only show you the temperature while you do those things.

Related

• Sudden FPS drops in games
• PC restarts or shuts down while gaming
• Gaming laptop slower than it should be
• GPU not at 100%