What is thermal throttling? Why phones slow down under sustained load

Every calculation a processor performs moves electrical charge around billions of tiny transistors, and almost all of that energy ends up as heat. The faster the chip runs, the more heat it produces. The relationship is not linear: the power a chip draws rises with its clock frequency and roughly with the square of its supply voltage, and higher frequencies usually need higher voltages. The last few steps of extra speed therefore cost far more power, and far more heat, than the first.

A desktop computer deals with this using heatsinks and fans. A phone has neither in most cases. Heat has to spread from the chip through internal layers, into the frame, the back panel and the screen, and from there into the surrounding air and your hand. That passive path works well for short bursts, but it can only move a limited amount of heat per second. How much depends on the phone's design, the room temperature and even whether the phone is in a thick case.

A modern phone is built around a system on a chip (SoC): the CPU cores, the GPU and many other components sit on one piece of silicon and share one power and thermal budget. Temperature sensors around the chip and the device report to the operating system and firmware, which constantly adjust each part's frequency and voltage. This is called dynamic voltage and frequency scaling.

When temperatures approach a limit, the system responds in a few typical ways:

• Lowering the maximum clock speed of the CPU cores, the GPU or both.
• Shifting work from the fastest cores to more efficient ones.
• Reducing other loads, such as screen brightness or charging speed, on some devices.

These limits exist for good reasons. Silicon and batteries age faster when they run hot, and a phone that becomes uncomfortable to hold is a problem in itself. Throttling keeps the device inside its safe operating range. It is designed behaviour, not a sign that something is broken.

Most everyday tasks are short: opening an app, loading a web page, taking a photo. Phones are tuned to run very fast for these bursts because the device starts cool and its own mass soaks up the first wave of heat. For a few seconds the chip can draw much more power than the phone could get rid of continuously.

Under continuous load, such as a long gaming session, a video export or a benchmark, heat keeps accumulating. Temperatures climb until the phone reaches a balance where the heat it produces matches the heat it can shed. The power level it can hold at that point is its sustainable power, and it is often well below the peak. Performance follows the same curve: high at first, then settling to a lower, steadier level.

A phone's peak speed tells you what it can do for a moment. Its sustained speed tells you what it can keep doing.

A test that finishes in a few seconds mostly measures burst performance. The phone never gets warm enough for throttling to kick in, so the result can look better than what you will experience during longer tasks.

A timed run of about a minute gives temperatures time to rise and any throttling time to show up in the work actually completed. A minute won't capture every long-session effect, because a phone can keep warming during an hour of gaming, but it is long enough to show whether performance holds steady or starts to fall as the temperature climbs.

This is how the MobiBench CPU stress test works. It runs computationally intensive work for about 60 seconds, and the CPU score is calculated from the work the device completes over that run, not looked up from the device name or brand. A phone that throttles heavily completes less work in the second half of the run, and the score reflects that.

Live graphs make throttling visible. MobiBench shows live CPU usage and temperature during the CPU stress test. Here is what common patterns usually mean.

A steady climb that levels off

Temperature rises quickly at first, then flattens. The phone is approaching its thermal balance. If the curve flattens well below the point where the phone feels hot, it is handling the load comfortably.

High usage, but less work done

CPU usage measures how busy the cores are, not how fast they are running. During throttling, usage can stay near the top of the graph while clock speeds drop, so the same busy cores complete less work per second. That is why a score based on completed work is more revealing than a usage graph on its own.

The starting temperature

Note where the temperature starts. A phone that begins a test already warm, from a previous run, charging or sitting in the sun, has less headroom before throttling begins, and its result will usually be lower.

No. Throttling is the mechanism that protects your phone. Running a demanding test will make the device warm, which is why MobiBench shows a warning about device load and heating before the CPU stress test starts. Stop a test if the phone becomes uncomfortably hot, and let it cool down before running another.

If you want results you can compare over time, see How to run a fair phone benchmark for a simple checklist.