상세정보
학술대회
AB9: 인공지능 추론 가속을 위한 프로세서
Cited - time in 
Download 12 time
Share 
Download 12 time
Share
- 발행일
- 202011
- 출처
- 대한전자공학회 학술 대회 (추계) 2020, pp.926-928
- 협약과제
- 20HS1900, 인공지능프로세서 전문연구실, 권영수
- 초록
- We present AB9, a neural processor for inference acceleration. AB9 consists of a systolic tensor core (STC) neural network accelerator designed to accelerate artificial intelligence applications by exploiting the data reuse and parallelism characteristics inherent in neural networks while providing fast access to large onchip memory. Complementing the hardware is an intuitive and userfriendly development environment that includes a simulator and an implementation flow that provides a high degree of programmability with a short development time. Along with a 40TFLOP STC that includes 32k arithmetic units and over 36？MB of onchip SRAM, our baseline implementation of AB9 consists of a 1GHz quadcore setup with other various industrystandard peripheral intellectual properties. The acceleration performance and power efficiency were evaluated using YOLOv2, and the results show that AB9 has superior performance and power efficiency to that of a generalpurpose graphics processing unit implementation. AB9 has been taped out in the TSMC 28nm process with a chip size of 17？×？23 mm2. Delivery is expected later this year.
- KSP 제안 키워드
- Acceleration performance, Development environment, Development time, Graphic Processing Unit(GPU), High degree, Intellectual property(IP), Network Accelerator, Neural processor, Power Efficiency, artificial intelligence applications, data reuse