A layer-wise frequency scaling for a neural processing unit

Dynamic voltage frequency scaling (DVFS) has been widely adopted for runtime power management of various processing units. In the case of neural processing units (NPUs), power management of neural network applications is required to adjust the frequency and voltage every layer to consider the power...

Full description

Saved in:
Bibliographic Details
Published in:ETRI journal 2022-10, Vol.44 (5), p.849-858
Main Authors: Chung, Jaehoon, Kim, HyunMi, Shin, Kyoungseon, Lyuh, Chun-Gi, Cho, Yong Cheol Peter, Han, Jinho, Kwon, Youngsu, Gong, Young-Ho, Chung, Sung Woo
Format: Article
Language:Korean
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Dynamic voltage frequency scaling (DVFS) has been widely adopted for runtime power management of various processing units. In the case of neural processing units (NPUs), power management of neural network applications is required to adjust the frequency and voltage every layer to consider the power behavior and performance of each layer. Unfortunately, DVFS is inappropriate for layer-wise run-time power management of NPUs due to the long latency of voltage scaling compared with each layer execution time. Because the frequency scaling is fast enough to keep up with each layer, we propose a layerwise dynamic frequency scaling (DFS) technique for an NPU. Our proposed DFS exploits the highest frequency under the power limit of an NPU for each layer. To determine the highest allowable frequency, we build a power model to predict the power consumption of an NPU based on a real measurement on the fabricated NPU. Our evaluation results show that our proposed DFS improves frame per second (FPS) by 33% and saves energy by 14% on average, compared with DVFS.
ISSN:1225-6463
2233-7326