Thermal Throttling: When Heat Becomes Your Bottleneck

A hot PC is not automatically a slow PC. Thermal throttling starts at one specific moment: when a chip cuts its own clock speed because a temperature sensor reached the limit the chip was designed for. Until then, a high reading is just a number. After it, heat is your bottleneck.

Quick Answer: What Is a Normal GPU and CPU Temperature?

Most graphics cards run at roughly 65–85°C at the GPU core while gaming and about 30–50°C at idle. The real ceiling is the maximum your specific card is rated for. NVIDIA lists it on each GeForce spec page (90°C for the RTX 4090), and AMD has said hotspot (junction) readings up to 110°C are within spec on Radeon cards. For CPUs, the limit is the chip’s Tjmax, for example 95°C on Ryzen 7000 and 89°C on Ryzen 7000 X3D models. A temperature only becomes a problem when it reaches that limit and the clock speed drops.

What Is Thermal Throttling?

Every CPU and GPU has a maximum safe temperature, called Tjmax on a CPU and the maximum GPU or junction temperature on a graphics card. When a sensor reaches it, the firmware lowers the clock speed, and often the voltage, until the temperature falls back. Less clock speed means fewer frames, so the frame rate drops even though nothing is broken. That drop is the thermal bottleneck. The temperature reading is only the clue that points to it.

It can look exactly like a hardware bottleneck in an FPS counter. The difference is the clock: if the clock stays at its normal boost speed and the frame rate is still low, heat is not the cause, and the CPU vs GPU bottleneck guide is the better place to look.

Ideal GPU Temperature While Gaming

There is no single ideal number for every card, but there are useful reference points:

GPU reading	What it usually means
30–50°C at idle	Normal for a desktop card with fans at low speed or stopped
65–85°C core while gaming	The range most hardware guides describe as normal under load
Near the card’s rated maximum	Watch the clock. If it holds, the card is fine; if it drops, it is throttling
Hotspot 10–20°C above the core	Expected, because the hotspot is the hottest point on the die
Hotspot far above the core, or core pinned at the maximum	Points to a cooling problem: dust, poor mounting, or dried thermal paste

Check your card’s own limit: NVIDIA publishes a “Maximum GPU Temperature” on the specification page of each GeForce card. The RTX 4090 is rated at 90°C, while the RTX 3090 Ti was rated at 93°C. Partner cards from ASUS, MSI, Gigabyte and others use the same GPU limit but different coolers, so two cards with the same chip can run at quite different temperatures.

GPU Temperature vs Hotspot Temperature

Monitoring tools often show two GPU readings. The GPU or edge temperature is an average across the chip. The hotspot (AMD calls it junction temperature) is the single hottest sensor on the die, so it always reads higher. In 2019 AMD stated that junction temperatures up to 110°C during typical gaming are expected and within spec on its Radeon RX 5700 cards.

A high hotspot alone is not a fault. A gap between core and hotspot that keeps growing over time usually is. When some reference Radeon RX 7900 XTX cards hit 110°C and throttled in 2022, AMD traced it to a problem with the cooler on those cards, not to normal behavior.

Ideal CPU Temperature While Gaming

Desktop CPUs are designed to boost until they reach a power or temperature limit, so a high reading in a heavy all-core load is often normal. Gaming loads the CPU less than a rendering benchmark, so gaming temperatures usually sit below the limit. What matters is the gap to your chip’s Tjmax:

CPU family	Maximum temperature (Tjmax)
AMD Ryzen 7000 (non-X3D)	95°C. AMD designed these chips to run at or near it under heavy load
AMD Ryzen 7000 X3D (7800X3D, 7900X3D, 7950X3D)	89°C
Intel Core 13th and 14th gen desktop (e.g. i9-14900K)	Commonly listed as 100°C. Check Intel’s spec page for your exact model

A reading of 85°C means very different things on a chip limited at 89°C and on one limited at 100°C. Compare your temperature to your own CPU’s limit, not to a number from a forum post about a different processor. Short spikes while a game compiles shaders or loads a map are normal; a temperature pinned at the limit with clocks that never recover is throttling.

Why Is My GPU So Hot?

• Dust in the cooler or case filters: the most common cause, and the cheapest to fix.
• Poor case airflow: too few intake fans, a front panel that blocks air, or a case pushed against a wall.
• A quiet fan curve: some cards keep their fans slow to stay silent, letting the temperature climb.
• Uncapped frame rate: a menu or old game running at hundreds of FPS keeps the GPU at full load for no visible benefit.
• Dried thermal paste or pads: common on cards several years old; the hotspot gap grows first.
• A warm room: the cooler can only get the chip a certain amount below the air temperature around it.

Why Is My CPU Overheating?

• A cooler too small for the chip: high-end CPUs that boost to their limit need a cooler sized for their power draw.
• Bad cooler mounting: uneven pressure or a forgotten protective film on the cooler base.
• Motherboard power settings: some boards ship with power limits far above the CPU’s defaults, so the chip runs hotter for little gain.
• Background load: a CPU near 100% while gaming runs hotter than it needs to; close what you don’t need.
• Dust and airflow: the same causes as on the GPU, especially on air coolers with dense fins.

How to Check for Throttling

HWiNFO: shows CPU thermal-throttling flags as “Yes/No”, plus GPU temperature and hotspot. Check the core clocks in the same window.

MSI Afterburner: its on-screen display can graph GPU temperature, hotspot (if the card exposes it) and GPU clock while you play. A sudden drop in the clock graph while the temperature line sits flat at its highest point is what throttling looks like.

A quick airflow test: run the same game scene twice, once normally and once with the case side panel off for a minute. If the clock recovers with the panel off, case airflow is the problem. If nothing changes, look at the cooler on the chip itself.

Laptop Throttling

Laptops hit both temperature limits and power limits, and the two are easy to confuse. A thin laptop can lower its clocks because its power budget is used up while the temperature still looks moderate. Cleaning the fans and keeping the vents clear fixes the temperature side; it does not change a power limit set by the manufacturer. Play on a hard, flat surface, use the performance mode in the manufacturer’s software, and judge the result by the clock speed, not by comparing your temperature with a desktop tower.

How to Fix Thermal Throttling, in Order

1. Clean and unblock: remove dust from filters, fans and heatsinks, and give the case room to breathe.
2. Raise the fan curve: a slightly louder GPU or case fan profile often removes throttling completely.
3. Cap the frame rate: limit FPS to your monitor’s refresh rate so the GPU isn’t working for frames you can’t see.
4. Undervolt: running the same clock at a lower voltage cuts heat and power. It is a tuning step, not something every PC needs.
5. Re-paste and re-seat the cooler: worthwhile when the hotspot gap is large or the hardware is several years old.
6. Upgrade the cooler or case: the last step, once the clock graph shows the current setup can’t hold boost in your games.

Once the clocks stay up, run the FPS calculator again: without a thermal limit, you should see the frame rate your parts are capable of. The bottleneck calculator describes how well your parts are matched, but it can’t see the temperature inside your case, so heat problems never show up in its result.

Frequently Asked Questions