Tensor Processing Unit TPU is a kind of application-specific integrated circuit ASIC developed specifically for machine learning. Compared with traditional GPU, its design is very clear. It mainly improves performance by reducing calculation precision and does not have hardware raster and texture mapping units.

The TPU chip is designed specifically for Google's TensorFlow, but Google currently still uses CPUs and GPUs for other types of machine learning. In addition to TPU, there is also the AI Accelerator, which is mainly used in the embedded and robotics fields.

References

【1】https://zh.wikipedia.org/wiki/TensorProcessor