Why does my FPS drop from 70 to 30 and then recover?

That cycle is the thermal loop drawn in numbers: performance falls at the limit, the cooler catches up during the low period, boost returns, and the loop repeats. A sawtooth curve in a sustained test is the same signature. The fix is the cooling ladder, not game settings.

Can capping FPS stop the drops?

Often, and it is a legitimate fix: capping at a rate the machine can sustain (say 60 instead of an uncapped 90) produces less heat, which can keep it below the throttle point entirely. A steady 60 beats a 90-to-35 sawtooth in every way that matters to hands.

Is it my CPU or GPU throttling?

In games, usually the GPU first on gaming laptops and the CPU first on thin machines, but they share a heatsink in most laptops, so one's heat becomes the other's throttle. A two-stage measurement (CPU load, then GPU load) shows which side gives first on yours.

Problems · Under load

Why does my FPS drop after 10 minutes of gaming?

FPS that starts high and falls after five to fifteen minutes is thermal in the overwhelming majority of cases. Games hold the processor and graphics card at full load together; the machine warms to its limit and steps performance down to protect itself. The timing is the fingerprint: broken hardware is bad immediately, heat needs minutes.

The pattern from the forums, 70–80 FPS falling to 30, recovering after a break, then falling again, is a thermal cycle described almost perfectly.

First, notice when it happens

From the first minute

Low FPS from the first minute is a different article: settings, drivers, the wrong GPU selected, or a machine below the game's weight. This page is about decline.

Only after minutes of use

Decline over minutes is accumulation, of heat. The only real alternatives are a power budget expiring (a clean step at a consistent time) or something scheduled arriving mid-session (check what else began running).

The causes, ranked

1. GPU or CPU thermal throttling

most common

How to confirm it

Gradual decline, loud fans, hot exhaust, worse in warm rooms, recovery after a rest. In-game overlays or HWiNFO show clocks stepping down as temperatures plateau at the limit.

The fix

The ladder, cheapest first: hard surface and clear vents (free), fan clean-out (cheap), fresh thermal paste (modest, transformative on machines past two years), undervolting where the platform allows it (free, effective, reversible), raised rear for airflow. Each rung moves the throttle later and the settled FPS higher.

2. The boost budget expired, by design

common

How to confirm it

A sharp step at a consistent time rather than a slide, identical in a cold room and a warm one. That is a configured power limit, not a fault.

The fix

Performance mode in the maker's app raises the budget on many gaming laptops (at the cost of noise). Beyond that, the post-step level is the machine's designed sustained gaming speed, the number reviews rarely print.

3. Heat soak: session after session starts warmer

common

How to confirm it

First session fine; each restart throttles sooner. The chassis and heatsink no longer return to cool between rounds.

The fix

Minutes of genuine idle between sessions, airflow around the machine, cooler room. Heat soak is cumulative; so is its fix.

4. Something else arrived mid-session

occasional

How to confirm it

The drop coincides with a download, a recording layer, a scan, or a game's own shader compilation, check the overlay's CPU graph and the task list at the moment of the drop.

The fix

Retime the intruder. This impostor matters because it wears heat's timing while having none of heat's fixes.

5. On battery, or on the wrong charger

occasional

How to confirm it

Gaming laptops unplugged run a fraction of their plugged power, and a USB-C travel charger cannot feed a gaming load even when plugged.

The fix

The full-wattage charger that shipped with the machine, in the wall, for every session that matters.

Measure it instead of guessing

The browser test measures exactly this decline, minus the game: sustained CPU throughput and a calibrated GPU load, sampled continuously. A machine whose test curve slides will slide in games, and the settled percentage predicts the FPS floor. Run it before and after a cleaning and the recovered frames are a number, not an impression.

Run the free test

What software cannot tell you here

Not your game's FPS, the game's own counter owns that number. We measure why the counter falls.
No temperatures or clock speeds; overlays and HWiNFO read those. The curve shows their consequence.

Asked, in people’s own words

Why does my FPS drop from 70 to 30 and then recover?: That cycle is the thermal loop drawn in numbers: performance falls at the limit, the cooler catches up during the low period, boost returns, and the loop repeats. A sawtooth curve in a sustained test is the same signature. The fix is the cooling ladder, not game settings.
Can capping FPS stop the drops?: Often, and it is a legitimate fix: capping at a rate the machine can sustain (say 60 instead of an uncapped 90) produces less heat, which can keep it below the throttle point entirely. A steady 60 beats a 90-to-35 sawtooth in every way that matters to hands.
Is it my CPU or GPU throttling?: In games, usually the GPU first on gaming laptops and the CPU first on thin machines, but they share a heatsink in most laptops, so one's heat becomes the other's throttle. A two-stage measurement (CPU load, then GPU load) shows which side gives first on yours.

Name the cause. Three and a half minutes, in the tab you have open.

Run the test