32TFLOPS function-safe programmable many-core AI processor32TFLOPS 기능안전 인공지능 프로세서

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 858
  • Download : 0
The proposed AI processor Architecture has high throughput for accelerating the neural network and re-duce the required external memory bandwidth for processing the neural network. For high throughput, the proposed super thread core (STC) includes 128 x 128 Nano Cores operating at 1.2GHz clock frequency. To reduce the required external memory bandwidth for reading all the kernel and feature data of the neural networks. The unified memory for kernel and feature data is proposed. Also, Using Smart Power-gating (Smart PG), the part of 16,384x NCs is powered and operating. The function-safe architecture is proposed for a fault tolerance system such as the electronics system for the autonomous car. The general-purpose processor core (GPP) is integrated with STC for controlling STC and processing AI algorithm and has the self-recovering cache and dynamic lockstep (DLS) function. the function-safe design has proved the fault performance has ASIL D of ISO26262 standard fault tolerance levels Therefore, the entire AI processor is fabricated in the 28nm CMOS process as a prototype chip. Its peak computing performance amounts to 32TFLOPS at 1GHz with 1.1V supply voltage. The measured 1.07TOPS/W energy efficiency is. a general-purpose processor for control with a function-safe design can have a 99.64% single-point fault tolerance rate.
Advisors
Yoo, Hoi-Junresearcher유회준researcher
Description
한국과학기술원 :전기및전자공학부,
Publisher
한국과학기술원
Issue Date
2020
Identifier
325007
Language
eng
Description

학위논문(박사) - 한국과학기술원 : 전기및전자공학부, 2020.2,[iv, 91 p. :]

Keywords

Artificial Intelligence Processor▼aFunction-Safe Design▼aMany-Core Architecture▼aArray Processor▼aNeural Network▼aDynamic Lockstep▼aSelf-recovering cache▼aAI Processor platform; 인공지능 프로세서▼a기능안전 설계▼a매니코어 아키텍처▼a어레이 프로세서▼a뉴럴 네트워크▼a동적 잠금 장치▼a자동 복구 캐시▼a인공지능 프로세서 플랫폼

URI
http://hdl.handle.net/10203/284189
Link
http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=909418&flag=dissertation
Appears in Collection
EE-Theses_Ph.D.(박사논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0