RAPL: Memory power estimation and capping

By Howard David, Eugene Gorbatov, U. Hanebutte, R. Khanna, C. Le. International Symposium on Low Power Electronics and Design, 2010

679 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Hardware and Architecture, Information Systems

View Paper: https://www.semanticscholar.org/paper/f5f838bba09fb5cb7261c26e2e0697d54f29a14d

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Hardware and Architecture

In Information Systems

Citations in the corpus

Memory power management via dynamic voltage/frequency scaling (2011, background): “in [5] describe RAPL (running average power limit), in which memory power is explicitly modelled and limited by throttling.”
Support Vector Regression (2020, neutral): “Validation: ALB-SVR was tested on a dataset used by David et al. (2010) and Stockman et al.”
Neural Cache: Bit-Serial In-Cache Acceleration of Deep Neural Networks (2018, methodology): “To measure execution power of the baseline, we use RAPL [27] for CPU power measurement and Nvidia-SMI [28] for GPU power measurement. widely-used SRAM.”
ACT: designing sustainable computer systems with an architectural carbon modeling tool (2022, background): “Over the last two decades, the computer architecture community has developed myriad tools and methods to quantify operational emissions by estimating andmeasuring utilization effectiveness [3, 15], energy consumption [20, 25, 53], and carbon intensity [1, 37].”
EnergyVis: Interactively Tracking and Exploring Energy Consumption for ML Models (2021, methodology): “[3]: it monitors CPU energy usage using Intel’s Running Average Power Limit interface [7], collects GPU usage through NVIDIA’s System Management interface, and is written in Python.”
Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning (2020, methodology): “We currently use Intel’s RAPL tool with the powercap interface (David et al., 2010) or Intel’s PowerGadget Tool 15 (depending on availability) to gather CPU/DRAM power draw and Nvidia’s nvidia-smi 16 for GPU power draw.”
Estimation of energy consumption in machine learning (2019, neutral): “[76] Instruction Current Analytical Energy Profiling CPU Model Yes [10] Hardware Analytical Analytical Power, Energy, Perf Simu data CPU Theoretical Yes [80] Hardware Analytical Transition-sensitive Energy/Power Simulation CPU, f units Theoretical – [44] Hardware Real Heuristics Power PMC, sim data ”