Automated Power and Latency Management in Heterogeneous 3D NoCs

Weldezion, Awet Yemane; Ebrahimi, Masoumeh; Daneshtalab, Masoud; Tenhunen, Hannu

doi:10.1145/2835512.2835517

Cited by 1 publication

(5 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…PTDBA [9], Cool-elevator [15], PTB3R [16] TLAR [5], TAAR [17], TTABR [6] Detailed operation [18] Considering the factors affecting the performance and reliability of 3D NoCs, it is not always possible to simultaneously achieve the required performance and thermal management through a heuristic approach [7]. Moreover, manually designing algorithm rules and strategies demands substantial engineering efforts.…”

Section: Related Workmentioning

confidence: 99%

“…However, excessive blocking occurs as the number of throttling nodes increases, resulting in a negative performance, owing to reduced routing flexibility. Detailed operation [18] Considering the factors affecting the performance and reliability of 3D NoCs, it is not always possible to simultaneously achieve the required performance and thermal management through a heuristic approach [7]. Moreover, manually designing algorithm rules and strategies demands substantial engineering efforts.…”

Section: Related Workmentioning

confidence: 99%

“…A routing scheme adopting this learning has a strong potential for significantly enhanced performance compared with conventional thermal-aware routing methods, by adapting the paths of packets to time-varying traffic loads in 3D NoCs [23]. Furthermore, the near-optimal solution of Q-learning is advantageous for constructing an optimal routing policy that considers performance and thermal management comprehensively in a 3D NoC, while considering more diverse factors than a 2D NoC [7]. However, the Q-table size overhead is problematic, as decisions are made using information regarding all paths for a specific node [23].…”

Section: Related Workmentioning

confidence: 99%

“…A Q-learning-based routing scheme alleviates packet transfer delays caused by network congestion in a 3D NoC, with the strength to directly generate optimal decision policies by learning runtime NoC states [7]. A learning packet containing traffic condition information is used to update the Q-table and the optimal output channel is determined based on the stored values.…”

Section: Q-learning-based Routing Algorithmmentioning

confidence: 99%

“…However, the increased thermal density in a 3D NoC eventually leads to saturation in performance and reliability, requiring additional cooling circuitry, such as a heat sink [7]. The stacked structure of the 3D NoC results in a longer heat dissipation path and a different inter-layer cooling efficiency [8], structure of the 3D NoC results in a longer heat dissipation path and a different inter-layer cooling efficiency [8], as shown in Figure 1.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Q-Function-Based Traffic- and Thermal-Aware Adaptive Routing for 3D Network-on-Chip

Lee

Han

2020

Electronics

View full text Add to dashboard Cite

Die-stacking technology is expanding the space diversity of on-chip communications by leveraging through-silicon-via (TSV) integration and wafer bonding. The 3D network-on-chip (NoC), a combination of die-stacking technology and systematic on-chip communication infrastructure, suffers from increased thermal density and unbalanced heat dissipation across multi-stacked layers, significantly affecting chip performance and reliability. Recent studies have focused on runtime thermal management (RTM) techniques for improving the heat distribution balance, but performance degradations, owing to RTM mechanisms and unbalanced inter-layer traffic distributions, remain unresolved. In this study, we present a Q-function-based traffic- and thermal-aware adaptive routing algorithm, utilizing a reinforcement machine learning technique that gradually incorporates updated information into an RTM-based 3D NoC routing path. The proposed algorithm initially collects deadlock-free directions, based on the RTM and topology information. Subsequently, Q-learning-based decision making (through the learning of regional traffic information) is deployed for performance improvement with more balanced inter-layer traffic. The simulation results show that the proposed routing algorithm can improve throughput by 14.0%–28.2%, with a 24.9% more balanced inter-layer traffic load and a 30.6% more distributed inter-layer thermal dissipation on average, compared with those obtained in previous studies of a 3D NoC with an 8 × 8 × 4 mesh topology.

show abstract