GAVINA: flexible aggressive undervolting for bit-serial mixed-precision DNN acceleration

TL;DR

GAVINA introduces a novel mixed-precision DNN accelerator that employs aggressive undervolting combined with bit-serial computation, significantly improving energy efficiency while maintaining accuracy.

Contribution

It presents GAVINA, a flexible architecture enabling aggressive undervolting and mixed-precision support, with an error model and substantial energy efficiency gains.

Findings

Energy efficiency up to 89 TOP/sW achieved

20% energy efficiency boost via undervolting

Negligible accuracy loss on ResNet-18

Abstract

Voltage overscaling, or undervolting, is an enticing approximate technique in the context of energy-efficient Deep Neural Network (DNN) acceleration, given the quadratic relationship between power and voltage. Nevertheless, its very high error rate has thwarted its general adoption. Moreover, recent undervolting accelerators rely on 8-bit arithmetic and cannot compete with state-of-the-art low-precision (<8b) architectures. To overcome these issues, we propose a new technique called Guarded Aggressive underVolting (GAV), which combines the ideas of undervolting and bit-serial computation to create a flexible approximation method based on aggressively lowering the supply voltage on a select number of least significant bit combinations. Based on this idea, we implement GAVINA (GAV mIxed-precisioN Accelerator), a novel architecture that supports arbitrary mixed precision and flexible undervolting, with an energy efficiency of up to 89 TOP/sW in its most aggressive configuration. By developing an error model of GAVINA, we show that GAV can achieve an energy efficiency boost of 20% via undervolting, with negligible accuracy degradation on ResNet-18.