Temperature-aware microarchitecture

DMCA

Temperature-aware microarchitecture (2003)

Cached

Download Links

Other Repositories/Bibliography

DBLP

by Kevin Skadron , Mircea R. Stan , Wei Huang , Sivakumar Velusamy , Karthik Sankaranarayanan , David Tarjan

Venue:	In Proceedings of the 30th Annual International Symposium on Computer Architecture
Citations:	478 - 52 self

BibTeX

@INPROCEEDINGS{Skadron03temperature-awaremicroarchitecture,
    author = {Kevin Skadron and Mircea R. Stan and Wei Huang and Sivakumar Velusamy and Karthik Sankaranarayanan and David Tarjan},
    title = {Temperature-aware microarchitecture},
    booktitle = {In Proceedings of the 30th Annual International Symposium on Computer Architecture},
    year = {2003},
    pages = {2--13}
}

OpenURL

Abstract

With power density and hence cooling costs rising exponentially, processor packaging can no longer be designed for the worst case, and there is an urgent need for runtime processor-level techniques that can regulate operating temperature when the package’s capacity is exceeded. Evaluating such techniques, however, requires a thermal model that is practical for architectural studies. This paper describes HotSpot, an accurate yet fast model based on an equivalent circuit of thermal resistances and capacitances that correspond to microarchitecture blocks and essential aspects of the thermal package. Validation was performed using finiteelement simulation. The paper also introduces several effective methods for dynamic thermal management (DTM): “temperaturetracking” frequency scaling, localized toggling, and migrating computation to spare hardware units. Modeling temperature at the microarchitecture level also shows that power metrics are poor predictors of temperature, and that sensor imprecision has a substantial impact on the performance of DTM. 1.

Keyphrases

temperature-aware microarchitecture package capacity microarchitecture level power metric equivalent circuit finiteelement simulation dynamic thermal management operating temperature poor predictor essential aspect thermal model hence cooling cost several effective method urgent need thermal package temperaturetracking frequency scaling power density architectural study runtime processor-level technique substantial impact thermal resistance hardware unit sensor imprecision

DMCA