Current browse context:
cs.NE
Change to browse by:
References & Citations
Computer Science > Neural and Evolutionary Computing
Title: ThUnderVolt: Enabling Aggressive Voltage Underscaling and Timing Error Resilience for Energy Efficient Deep Neural Network Accelerators
(Submitted on 11 Feb 2018 (v1), last revised 13 Mar 2018 (this version, v2))
Abstract: Hardware accelerators are being increasingly deployed to boost the performance and energy efficiency of deep neural network (DNN) inference. In this paper we propose Thundervolt, a new framework that enables aggressive voltage underscaling of high-performance DNN accelerators without compromising classification accuracy even in the presence of high timing error rates. Using post-synthesis timing simulations of a DNN accelerator modeled on the Google TPU, we show that Thundervolt enables between 34%-57% energy savings on state-of-the-art speech and image recognition benchmarks with less than 1% loss in classification accuracy and no performance loss. Further, we show that Thundervolt is synergistic with and can further increase the energy efficiency of commonly used run-time DNN pruning techniques like Zero-Skip.
Submission history
From: Jeff (Jun) Zhang [view email][v1] Sun, 11 Feb 2018 19:51:59 GMT (1659kb,D)
[v2] Tue, 13 Mar 2018 04:10:47 GMT (1659kb,D)
Link back to: arXiv, form interface, contact.