An integrated GPU power and performance model

By Sunpyo Hong, Hyesoon Kim. International Symposium on Computer Architecture, 2010

586 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Electrical and Electronic Engineering, Hardware and Architecture

View Paper: https://www.semanticscholar.org/paper/1d286a264b233125b681e522e8f5fed596a8608c

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 Electrical and Electronic Engineering

In Hardware and Architecture

Citations in the corpus

GPUWattch: enabling energy optimizations in GPGPUs (2013, methodology): “Weusean iterative process to continuously reﬁne the power model based on inaccuracies that we observe between the power model and the actual hardware power measurements.”
A Survey on Deep Learning (2018, background): “However, GPUs are notorious for their leakage currents, and this abstracts any plausible realization of the deep learning models on portable devices [63].”
Accel-Sim: An Extensible Simulation Framework for Validated GPU Modeling (2018, background): “Many previous works [4], [20], [21], [71], [74], [75] aim to estimate the execution time of GPGPU applications throughout source-level analysis, however they do not perform detailed performance modeling for architecture research.”
JointDNN: An Efficient Training and Inference Engine for Intelligent Mobile Cloud Computing Services (2018, background): “Industrial developers do not reveal the detailed hardware architecture specifications and the proprietary parallel computing architectures such as CUDA , therefore, the analytical approach could be quite challenging [20].”
Data Center Energy Consumption Modeling: A Survey (2016, background): “In another example for additive power models, Hong et al. described modeling power consumption of GPUs [154].”
A Scalable Near-Memory Architecture for Training Deep Neural Networks on Large In-Memory Datasets (2018, methodology): “Compared to theGPUpower analysis and model published in [26], NTX spends a larger fraction of power in the FPUs, namely 14 versus 4.”
Accelerometer: Understanding Acceleration Opportunities for Data Center Overheads at Hyperscale (2020, background): “model GPU execution time [56] and power requirements [57].”
Kernel Fusion: An Effective Method for Better Power Efficiency on Multithreaded GPU (2010, background): “The most recent study done by Hong [12] established a integrated performance and power model which could be used to predict the optimal number of active cores reaching the highest power efficiency.”
Understanding the Future of Energy Efficiency in Multi-Module GPUs (2019, methodology): “Commonly used GPU energy estimation frameworks adopt a bottom-up modeling approach [15], [24], where the energy cost of each key microarchitectural component is combined with the corresponding switching activities to determine the overall GPU energy consumption.”