BROWSE
Detail
Conference Paper
CPU Interconnects의 진화
Cited - time in 
Share 
Share
- Authors
- 정병권
- Issue Date
- 2021-11
- Citation
- 대한전자공학회 학술 대회 (추계) 2021, pp.729-733
- Publisher
- 대한전자공학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- Recently, high-performance computing systems are required for big data processing such as AI, image classification, voice recognition, medical diagnosis, and autonomous driving. Accordingly, collection, storage, processing, and analysis of big data in search engines, finance, and telecommunications, as well as handling multimedia big data, which has been previously easy to access and utilize, are increasingly required. In recent years, the size and type of big data that can be digitized and accepted are expanding further with the advancement of computing technology due to the rapid development of storage and memory semiconductor technology. As such faster data processing and a larger and more efficient memory approach are required, such data processing requires support for high bandwidth connecting processors and devices inside computers and connecting devices to devices. Currently, most I/O devices such as CPUs and GPUs and accelerators are connected through PCIe protocols, resulting in relatively low bandwidth and high delays, memory sharing and consistency problems between I/O and CPU.BRThe computer industry has been seeking more efficient protocols for computer connections to expand the connection of processors and devices inside computers since 2015. In this paper, we will deal with the evolution of the architecture surrounding the CPU.
- KSP Keywords
- Analysis of big data, Computing technology, Efficient protocols, High-performance computing(HPC), High-performance computing systems, I/O devices, Image Classification, Medical diagnosis, Memory approach, Memory sharing, Rapid development