System Architecture and Software Stack for GDDR6-AiM

This poster presents system architecture, software stack, and performance analysis for SK hynix's very first GDDR6-based processing-in-memory (PIM) product sample, called Accelerator-in-Memory (AiM).AiM is designed for the in-memory acceleration of matrix-vector product operations, which are co...

Full description

Saved in:

Bibliographic Details
Published in:	2022 IEEE Hot Chips 34 Symposium (HCS) pp. 1 - 25
Main Authors:	Kwon, Yongkee, Vladimir, Kornijcuk, Kim, Nahsung, Shin, Woojae, Won, Jongsoon, Lee, Minkyu, Joo, Hyunha, Choi, Haerang, Kim, Guhyun, An, Byeongju, Kim, Jeongbin, Lee, Jaewook, Kim, Ilkon, Park, Jaehan, Park, Chanwook, Song, Yosub, Yang, Byeongsu, Lee, Hyungdeok, Kim, Seho, Kwon, Daehan, Lee, Seongju, Kim, Kyuyoung, Oh, Sanghoon, Park, Joonhong, Hong, Gimoon, Ka, Dongyoon, Hwang, Kyudong, Park, Jeongje, Kang, Kyeongpil, Kim, Jungyeon, Jeon, Junyeol, Lee, Myeongjun, Shin, Minyoung, Shin, Minhwan, Cha, Jaekyung, Jung, Changson, Chang, Kijoon, Jeong, Chunseok, Lim, Euicheol, Park, Il, Chun, Junhyun, Hynix, Sk
Format:	Conference Proceeding
Language:	English
Published:	IEEE 21-08-2022
Subjects:	Bandwidth Computer architecture Hardware Machine learning Random access memory Software systems Systems architecture
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	This poster presents system architecture, software stack, and performance analysis for SK hynix's very first GDDR6-based processing-in-memory (PIM) product sample, called Accelerator-in-Memory (AiM).AiM is designed for the in-memory acceleration of matrix-vector product operations, which are commonly found in machine learning applications. The strength of AiM primarily comes from the two design factors, which are 1) all-bank operation support and 2) extended DRAM command set. All-bank operations allow AiM to fully utilize the abundant internal DRAM bandwidth, which makes it an attractive solution for memory-bound applications. The extended command set allows the host to address these new operations efficiently and provides a clean separation of concerns between the AiM architecture and its software stack design.We present a dedicated FPGA-based reference platform with a software stack, which is used to validate AiM design and evaluate its system-level performance. We also demonstrate FMC-based AiM extension cards that are compatible with the off-the-shelf FPGA boards and serve as an open research platform allowing potential collaborators and academic institutes to access our hardware and software systems.
ISSN:	2573-2048
DOI:	10.1109/HCS55958.2022.9895629