Experimental Characterization of Variation in Power Consumption for Processors of Different Generations

Abstract

Data centers are energy-hungry facilities. Building energy consumption predictive models for servers is one of the solutions to use efficiently the resources. However, physical experiments have shown that even under the same conditions, identical processors consume different amount of energy to complete the same task. While this manufacturing variability has been observed and studied before, there is lack of evidence supporting the hypotheses due to limited sampling data, especially from the thermal characteristics. In this article, we compare the power consumption among identical processors for two Intel processors series with the same TDP (Thermal Design Power) but from different generations. The observed power variation of the processors in newer generation is much greater than the older one. Then, we propose our hypotheses for the underlying causes and validate them under precisely controlled environmental conditions. The experimental results show that, with the increase of transistor densities, difference of thermal characteristics becomes larger among processors, which has non-negligible contribution to the variation of power consumption for modern processors. This observation reminds us of re-calibrating the precision of the current energy predictive models. The manufacturing variability has to be considered when building energy predictive models for homogeneous clusters.

Taming Energy Consumption Variations In Systems Benchmarking

“…The aim of this expriment is not to justify the evolution of the variation across CPU versions/generations, but to observe if the user can choose the best node to execute her experiments. Previous papers have discussed the evolution of the energy consumption variation across CPU generations and concluded that the variation is getting higher with the latest CPU generations [19,27], which makes measurements stability even worse. In this experiment, we therefore compare four different generations of CPU with the aim to evaluate the energy variation for each CPU and its correlation with the generation.…”

“…Figures 12 and 13 also highlight that applying the guidelines does not reduce the inter-nodes variation in all the cases. This variation can be up to 30 % in modern CPU [27]. However, taming the intra-node variation is a good strategy to identify more relevant mediums and medians, and then perform accurate comparisons between the nodes variation.…”

Is Hot IT a False Economy? An Analysis of Server and Data Center Energy Efficiency as Temperatures Rise

“…In 2018, Wang et al [30] observed a correlation between CPU temperature and server power by intermittently cooling a CPU in isolation. Later, they profiled a single server's power consumption under load and up to 45 • C to generate a temperature-aware model [31]. Both tests showed increased power consumption under hotter inlet temperatures or reduced cooling.…”

HDSAP: Heterogeneity-aware dynamic scheduling algorithm to improve performance of nano-scale many-Core processors for unpredictable workloads

“…The frequency variation in an 80-core processor within a single die in Intel's 65 nm technology is 28% at 1.2v and 59% at 0.8v voltage levels [8]. Furthermore, the amount of total power variation is about 30% among 30 identical processors, without the effect of frequency variation [9]. In 45-nm, there is a −33.5% to 81.7% difference in critical charge on a chip [10] that leads to variation in soft error rate (SER) among cores.…”