A System-Level Voltage/Frequency Scaling Characterization Framework for Multicore CPUs

• DOI: 10.48550/arxiv.2106.09975
• Publication Date: 2021-01-01
• Authors: George N. Papadim...; Manolis Kaliorakis; Athanasios Chatzi...; Dimitris Gizopoulos; Greg Favor; Kumar Sankaran; Shidhartha Das
• Keywords: Microprocessor; Frequency scaling; Multi-core processor; Characterization

Supply voltage scaling is one of the most effective techniques to reduce power consumption microprocessors. However, technology limitations such as aging and process variability enforce microprocessor designers apply pessimistic guardbands guarantee correct operation in field for any foreseeable workload. This worst-case design practice makes energy efficiency hard scale with evolution. Improving energy-efficiency requires identification chip margins through time-consuming comprehensive characterization its operational limits. Such a state-of-the-art multi-core CPUs fabricated aggressive technologies multi-parameter process, which statistically significant information. In this paper, we present an automated framework support system-level frequency Applied Micro's ARMv8-based multicore used X-Gene 2 micro-server family. The fully can provide fine-grained information system's state by monitoring abnormal behavior that may occur during reduced supply conditions. We also propose new metric quantify when it operates beyond nominal Our experimental results demonstrate potential uses identify limits improved efficiency.