NPU
An NPU, or neural processing unit, is a part of a chip built to run artificial intelligence tasks quickly while using little power.
On this page 5 sections
How it works
Artificial intelligence features rely on a very large number of simple calculations, mostly multiplying and adding numbers. A normal processor core handles these one after another, and a graphics processor handles many at once but uses a lot of power. An NPU is a dedicated block inside the same chip designed for exactly this kind of work, so it can run the calculations in parallel using far less energy.
What it is used for
- Camera features such as scene detection, noise reduction and background blur.
- Voice recognition, live captions and translation done on the device itself.
- Running small AI models without sending data to a remote server.
Other names
Some chip makers call this block an AI engine or AI accelerator, but the idea is the same.
What the numbers mean
NPU speed is often given in TOPS, meaning trillions of operations per second. The figure depends on the number format used in the test, so figures from different makers are not always comparable.
What it does not promise
An NPU does not make a device faster at everyday tasks such as opening apps. It helps only software written to use it, and large AI models may still run on a remote server.