table of Contents
Article Directory
Preface
To be added.
Edge computing and hardware vendors
Edge servers must support heterogeneous computing in order to meet multiple business demands and computing requirements for diverse data. The core of heterogeneous computing is multi-chip support, including: CPU (x86, ARM), GPU, NPU, NP, etc.
GPUs are widely used in video encoding and decoding, parallel computing, and manual computing. Typical manufacturers are NVIDIA, AMD, etc. NPU is a neural network processor that uses a data-driven parallel computing architecture. It has a wide range of applications in artificial intelligence and deep learning. Typical manufacturers include Cambrian, Shengteng, etc. Typical manufacturers are Broadcom, Marvell, etc.
The ARM-based CPU occupies most of the market share in the terminal field. With the continuous introduction of ARM high-performance cores, it can also meet the application in the server field. Especially in the field of edge computing, as the first entry point for data, the advantages of the ARM architecture in the terminal field can better realize end-to-side collaboration and cope with the diversity of massive data.
The edge computing server is compatible with PCIE, DDR and other basic hardware specifications in hardware to ensure the integrity of the hardware ecosystem.
The operating system is supported by mainstream CentOS, Kylin, EulerOS, etc., and is compatible with mainstream AI frameworks, such as TensorFlow, CUDA (Compute Unified Device Architecture), MindSpore, etc.
The article focuses on the analysis of server architecture composition, technical changes brought by 5G to servers (computing boundaries, scenarios, etc.), edge computing and cloud computing requirements for servers (edge servers, AI servers, and cloud servers), servers (white-branded machines, open source servers) , Hardware reconstruction and software definition) and server key vendors.