Scholarlink

This research aims to address the existing knowledge gap in predicting minimum operating voltage (Vmin) in microprocessor design through various machine-learning methods. As the demand for energy-efficient and high-performance microprocessors grow, accurate Vmin prediction becomes crucial. With a focus on Vmin search procedure and machine-learning integration, the research streamlines testing, cuts costs, prevents degradation of unit, and enhances the product lifespan in semiconductor manufacturing. This paper thoroughly examines the relationship between Intra-Die Variation (IDV), Leakage Current, and Vmin. This study introduces a predictive model for microprocessor design, emphasizing Machine Learning Algorithms-Linear Regression, Random Forest, and Gradient Boosting. Notably, Linear Regression emerges as the most effective, predicting a mere 2.681% voltage variation for high frequency. The proposed model, leveraging Linear Regression, proves resilient, reducing excess voltage and optimizing production processes.

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

Minimum Operating Voltage Prediction by Machine-Learning Algorithms in Microprocessor

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Niraj Chopade,
  • Madhuri A. Chaudhari,
  • Rohit Kumar,
  • Prasanna Onkar

摘要

This research aims to address the existing knowledge gap in predicting minimum operating voltage (Vmin) in microprocessor design through various machine-learning methods. As the demand for energy-efficient and high-performance microprocessors grow, accurate Vmin prediction becomes crucial. With a focus on Vmin search procedure and machine-learning integration, the research streamlines testing, cuts costs, prevents degradation of unit, and enhances the product lifespan in semiconductor manufacturing. This paper thoroughly examines the relationship between Intra-Die Variation (IDV), Leakage Current, and Vmin. This study introduces a predictive model for microprocessor design, emphasizing Machine Learning Algorithms-Linear Regression, Random Forest, and Gradient Boosting. Notably, Linear Regression emerges as the most effective, predicting a mere 2.681% voltage variation for high frequency. The proposed model, leveraging Linear Regression, proves resilient, reducing excess voltage and optimizing production processes.