Scholarlink

<p>Three-dimensional network-on-chip (3D NoC) architectures face critical thermal management challenges due to concentrated heat from vertically stacked dies, leading to hotspots and performance degradation. To address this, the adaptive multi-beltway thermal-aware routing (AMBTAR) algorithm, a novel proactive routing strategy, is proposed. AMBTAR integrates three distinct non-minimal routing modes, forward-directed beltway (FDB), reverse-directed intra-layer beltway (Intra-RDB), and reverse-directed inter-layer beltway (Inter-RDB), with traditional minimal routing. The selection is governed by a probabilistic mechanism that uses a cost function to dynamically balance thermal benefits against path length overhead. Experimental results demonstrate that the algorithm’s ability to intelligently distribute traffic leads to significant benefits, including up to a 37.79% reduction in temperature standard deviation, up to 47.3% lower average packet latency, and 12.9% higher throughput compared to baseline algorithms. Furthermore, this balanced approach results in notable energy savings, with consumption reduced by up to 12.0%. By significantly increasing path diversity and proactively managing thermal load, AMBTAR enhances overall network performance and reliability in thermally constrained 3D NoC environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive multi-beltway thermal-aware routing algorithm for 3D NoC system

  • 期刊论文

  • 发表日期：

    2025年7月1日

查看全文
  • Mohammad Reza Dadmand,
  • Razieh Farazkish,
  • Akram Reza,
  • Roshanak Rafiei Nazari,
  • Reza Faghih Mirzaee

摘要

Three-dimensional network-on-chip (3D NoC) architectures face critical thermal management challenges due to concentrated heat from vertically stacked dies, leading to hotspots and performance degradation. To address this, the adaptive multi-beltway thermal-aware routing (AMBTAR) algorithm, a novel proactive routing strategy, is proposed. AMBTAR integrates three distinct non-minimal routing modes, forward-directed beltway (FDB), reverse-directed intra-layer beltway (Intra-RDB), and reverse-directed inter-layer beltway (Inter-RDB), with traditional minimal routing. The selection is governed by a probabilistic mechanism that uses a cost function to dynamically balance thermal benefits against path length overhead. Experimental results demonstrate that the algorithm’s ability to intelligently distribute traffic leads to significant benefits, including up to a 37.79% reduction in temperature standard deviation, up to 47.3% lower average packet latency, and 12.9% higher throughput compared to baseline algorithms. Furthermore, this balanced approach results in notable energy savings, with consumption reduced by up to 12.0%. By significantly increasing path diversity and proactively managing thermal load, AMBTAR enhances overall network performance and reliability in thermally constrained 3D NoC environments.