Reducing Hibernation Energy and Degradation in Bipolar ReRAM-Based Non-Volatile Processors

ReRAM-based Non-Volatile Flip-Flops (NVFFs) enable instant hibernation and zero-leakage sleep modes that are highly desired in energy harvesting and frequently-off Non-Volatile Processors (NVPs). However, their high store energy demand and rapid degradation due to excess electrical stress and unnece...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on nanotechnology 2019, Vol.18, p.657-669
Hauptverfasser: Biglari, Mehrdad, Lieske, Tobias, Fey, Dietmar
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:ReRAM-based Non-Volatile Flip-Flops (NVFFs) enable instant hibernation and zero-leakage sleep modes that are highly desired in energy harvesting and frequently-off Non-Volatile Processors (NVPs). However, their high store energy demand and rapid degradation due to excess electrical stress and unnecessary writes during store time remain open challenges for the deployment of ReRAM-based NVP. To address these concerns, this paper presents two energy-efficient and low-degradation bipolar ReRAM-based NVFFs, Hypnos, and Morpheus. In Hypnos, we reduce the ReRAM electrical stress during set operation while keeping the introduced area overhead at a minimum. In Morpheus, a write-termination circuit reduces the store energy demand and excess electrical stress even further at the cost of an affordable area overhead. Both NVFFs feature run-time tunable resistive states to enable the online adjustment of their tradeoff among endurance, retention, energy demand, and restore success rate. Experimental results demonstrate the enhanced energy efficiency and endurance properties of the proposed NVFFs. We further develop NVPs based on Hypnos and Morpheus and investigate the performance of the resulting NVPs in terms of effective processing frequency, hibernation energy, and degradation in a case study on battery-less and battery-powered Internet-of-Things (IoT) devices. In particular, we show that Hypnos's stress reduction mechanism reduces the NVP's set degradation by 78 %, whereas Morpheus's write-termination mechanism alleviates the remaining NVP degradation by 66% (set) and 88 % (reset).
ISSN:1536-125X
1941-0085
DOI:10.1109/TNANO.2019.2922363