Many-cores systems on chip provide the highest performance scaling potential due to the massive parallelism, but they suffer from thermal issues due to their high power densities. Thermal sensors and feedback strategies are used to reduce these threats but sensor accuracy directly impact control performance. In this paper we propose a novel technique to calibrate thermal sensors. Our approach can be applied to general multi-core platforms since it combines stress patterns and least-square fitting to perform thermal sensor characterization directly on the target device. We experimentally validate our approach on the Single Chip Cloud (SCC) prototype by Intel.
A System Level Approach to Multi-core Thermal Sensors Calibration / Bartolini A.; Sadri M.; Beneventi F.; Cacciari M.; Tilli A.; Benini L.. - STAMPA. - 6951:(2011), pp. 22-31. (Intervento presentato al convegno PATMOS 2011 tenutosi a Madrid, Spain nel September 26-29, 2011) [10.1007/978-3-642-24154-3_3].
A System Level Approach to Multi-core Thermal Sensors Calibration
BARTOLINI, ANDREA;SADRI, MOHAMMADSADEGH;BENEVENTI, FRANCESCO;CACCIARI, MATTEO;TILLI, ANDREA;BENINI, LUCA
2011
Abstract
Many-cores systems on chip provide the highest performance scaling potential due to the massive parallelism, but they suffer from thermal issues due to their high power densities. Thermal sensors and feedback strategies are used to reduce these threats but sensor accuracy directly impact control performance. In this paper we propose a novel technique to calibrate thermal sensors. Our approach can be applied to general multi-core platforms since it combines stress patterns and least-square fitting to perform thermal sensor characterization directly on the target device. We experimentally validate our approach on the Single Chip Cloud (SCC) prototype by Intel.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
