16 Years of SPEC Power: An Analysis of x86 Energy Efficiency Trends

Hannes Tröpgen,Robert Schöne,Thomas Ilsche,Daniel Hackenberg

DOI: https://doi.org/10.1109/CLUSTERWorkshops61563.2024.00020

2024-11-11

Abstract:The SPEC Power benchmark offers valuable insights into the energy efficiency of server systems, allowing comparisons across various hardware and software configurations. Benchmark results are publicly available for hundreds of systems from different vendors, published since 2007. We leverage this data to perform an analysis of trends in x86 server systems, focusing on power consumption, energy efficiency, energy proportionality and idle power consumption. Through this analysis, we aim to provide a clearer understanding of how server energy efficiency has evolved and the factors influencing these changes.

Hardware Architecture

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to analyze the energy - efficiency trends of x86 server systems over the past 16 years. Specifically, the author uses SPEC Power benchmark test data to explore the following aspects:

1. **Power Consumption and Energy - Efficiency Trends**: By analyzing SPEC Power benchmark test results, study the power consumption and energy - efficiency change trends of x86 server systems at different load levels. This includes the changes in maximum power consumption, partial - load power consumption, and energy efficiency (in units of ssj_ops/W).

2. **Energy Proportionality**: Evaluate the energy - use efficiency of servers at different load levels, that is, energy proportionality. This refers to whether the system can effectively adjust its energy consumption at different loads to achieve higher energy efficiency.

3. **Idle Power Consumption Trends**: Analyze the power - consumption change trends of servers in the idle state. Idle power consumption refers to the power consumption when the server is not processing any tasks, which is especially important for high - utilization high - performance computing (HPC) systems.

4. **Influence of Manufacturers and Technologies**: Compare the influence of different CPU manufacturers (such as Intel and AMD) and different technologies (such as dynamic voltage and frequency scaling DVFS, core C - states, etc.) on energy efficiency. In particular, the contribution of AMD's EPYC processor to energy - efficiency improvement since 2017.

5. **Future Development Trends**: Explore the current challenges faced by energy - efficiency optimization and look forward to possible future development directions, such as the application of the new SPECpowerNext benchmark test suite.

Through these analyses, the paper aims to provide a clear understanding, revealing the trends in the evolution of energy efficiency in x86 server systems over the past 16 years and their influencing factors. This not only helps hardware manufacturers improve product design, but also provides valuable reference information for data - center operators.

### Key Formulas
- Energy Efficiency: \[ \text{Energy Efficiency} = \frac{\text{Performance (ssj\_ops)}}{\text{Power Consumption (W)}} \]
- Overall Efficiency: \[ \text{Overall Efficiency} = \frac{\sum \text{ssj\_ops}}{\sum P} \]
where \(\sum P\) represents the average power at all load levels (including the active - idle state).

### Summary
The paper, through in - depth analysis of a large number of SPEC Power benchmark test results, reveals the significant progress and challenges in energy efficiency of x86 server systems, providing important guidance for future energy - efficiency optimization.