Publication
HPCA 2019
Conference paper

Resilient low voltage accelerators for high energy efficiency

View publication

Abstract

Low voltage architecture and design are key enablers of high throughput per watt in heterogeneous, accelerator-rich manycore designs. However, such low voltage operation poses significant challenges because of difficulties in achieving reliable functionality of on-chip memories, particularly SRAMs at these design points. In this paper, we present a technique of low-voltage neural network acceleration, where the embedded SRAM architecture is equipped with a novel applicationaware supply voltage boosting capability. This technique mitigates low-voltage induced failures, while enabling Very low voltage (VLV)1 operation during most of the application run, resulting in substantial improvement in net energy efficiency. We present a framework to evaluate the impact of low-voltage SRAM errors on machine learning applications and characterize trade-offs between output inference accuracy and energy efficiency in our application-programmable supply boosted SRAM architecture. Using the proposed technique we push the limits on the minimum operable voltage (Vmin) for the desired output quality. As a proof of concept, we demonstrate these techniques on Dante, a Deep Neural Network (DNN) accelerator chip taped out in state-of-the art 14nm technology.

Date

26 Mar 2019

Publication

HPCA 2019

Authors

Share