• Complex
  • Title
  • Keyword
  • Abstract
  • Scholars
  • Journal
  • ISSN
  • Conference
搜索

Author:

Fang, Juan (Fang, Juan.) (Scholars:方娟) | Cai, Min (Cai, Min.) | Leng, Zhenyu (Leng, Zhenyu.) | Liu, Sitong (Liu, Sitong.)

Indexed by:

CPCI-S

Abstract:

Dark silicon leads the integration development of multi-core chip to a standstill, which is mainly due to on-chip power constraints. NoC power occupies a significant portion of system power. Therefore, reducing the NoC power can reduce the dark silicon degree and reduce the performance degradation caused by the dark silicon. In this paper, we firstly analyzed the NoC power variation on the dark silicon degree, which has not been done before, and proposed a heterogeneous NoC scheme to reduce the dark silicon degree. By evaluating the power of homogeneous NoCs and heterogeneous NoCs separately under different power budgets, we verified that when NoCs adopt this heterogeneous scheme, most of the workloads would get one more core power saving from the dark silicon. In other words, at least one more core could be lighted on and the dark silicon degree can be reduced.

Keyword:

NoCs networks-on-chip dark silicon

Author Community:

  • [ 1 ] [Fang, Juan]Beijing Univ Technol, Coll Comp Sci, Beijing, Peoples R China
  • [ 2 ] [Cai, Min]Beijing Univ Technol, Coll Comp Sci, Beijing, Peoples R China
  • [ 3 ] [Liu, Sitong]Beijing Univ Technol, Coll Comp Sci, Beijing, Peoples R China
  • [ 4 ] [Leng, Zhenyu]Chinese Acad Sci, Inst Comp Technol, State Key Lab Comp Architecture, Beijing, Peoples R China

Reprint Author's Address:

  • 方娟

    [Fang, Juan]Beijing Univ Technol, Coll Comp Sci, Beijing, Peoples R China

Show more details

Related Keywords:

Related Article:

Source :

2016 INTERNATIONAL CONFERENCE ON COMPUTATIONAL SCIENCE & COMPUTATIONAL INTELLIGENCE (CSCI)

Year: 2016

Page: 590-595

Language: English

Cited Count:

WoS CC Cited Count: 0

SCOPUS Cited Count:

ESI Highly Cited Papers on the List: 0 Unfold All

WanFang Cited Count:

Chinese Cited Count:

30 Days PV: 0

Affiliated Colleges:

Online/Total:776/6464890
Address:BJUT Library(100 Pingleyuan,Chaoyang District,Beijing 100124, China Post Code:100124) Contact Us:010-67392185
Copyright:BJUT Library Technical Support:Beijing Aegean Software Co., Ltd.