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

中文翻译：

---

多核 CPU 的系统级电压/频率缩放特性框架

电源电压调节是降低微处理器功耗的最有效技术之一。然而，老化和工艺可变性等技术限制迫使微处理器设计人员应用悲观电压保护带，以确保在任何可预见的工作负载下都能在现场正确运行。这种最坏情况的设计实践使得能源效率难以随着技术的发展而扩展。提高能效需要通过对其操作限制进行耗时且全面的表征来确定芯片设计余量。采用先进技术制造的最先进多核 CPU 的这种表征是一个多参数过程，需要具有统计意义的信息。在本文中，我们提出了一个自动化框架，以支持 X-Gene 2 微服务器系列中使用的 Applied Micro 最先进的基于 ARMv8 的多核 CPU 的系统级电压和频率缩放特性。完全自动化的框架可以通过监控在降低的电源电压条件下可能发生的任何异常行为来提供系统状态的细粒度信息。我们还提出了一个新指标来量化微处理器在超出标称条件下运行时的行为。我们的实验结果证明了表征框架的潜在用途，以确定提高能源效率的操作限制。完全自动化的框架可以通过监控在降低的电源电压条件下可能发生的任何异常行为来提供系统状态的细粒度信息。我们还提出了一个新指标来量化微处理器在超出标称条件下运行时的行为。我们的实验结果证明了表征框架的潜在用途，以确定提高能源效率的操作限制。完全自动化的框架可以通过监控在降低的电源电压条件下可能发生的任何异常行为来提供系统状态的细粒度信息。我们还提出了一个新指标来量化微处理器在超出标称条件下运行时的行为。我们的实验结果证明了表征框架的潜在用途，以确定提高能源效率的操作限制。