News subscription required to download the report. Please sign in if you wish to continue.
With the rise of edge computing, network processing units (NPUs) will be one of the key developing directions for AI accelerator chips.
Abstract
In line with the rise of edge computing, the computing tasks are gradually decentralized to edge servers and terminal devices to respond to the requirements for latency, energy-saving, and information security under different scenarios, with the expectation that more highly specialized edge AI accelerator chips will likely to enter into the market and that neural network processing unit (NPU) will be one of the crucial developing directions that will achieve high price/performance ratio regarding low power consumption and higher computing capability and enable flexible adjustments in various edge computing environments.
DIGITIMES Research believes that the NPU will bring new choices for AI accelerator chips in edge devices. Compared to general-purpose computing on graphics processing units (GPGPU), NPU has power consumption and price/performance ratio advantage, making it suitable to handle AI computing tasks in edge computing nodes.
However, due to the high specialization of NPU, it can only execute AI algorithms under specific frameworks, and the software development environment is relatively closed, posing a significant challenge to the development of the NPU industry.
Table of contents
