Variation-aware thermal characterization and management of multi-core architectures

Abstract

The accuracy and efficiency of dynamic power and thermal management are both affected by the increased levels of on-chip variation, mainly because dynamic thermal management schemes are oblivious to the variation characteristics of the underlying hardware. We propose a technique that utilizes the existing on-chip sensor infrastructure to improve the inherent thermal imbalances among different cores in a multicore architecture. Thermal sensor readings are compiled to generate an on-chip variation map, which is provided to the system power/thermal management to effectively manage the existing on-chip variation. Experimental analysis based on live measurements on a special test-chip shows reduced on-chip heating with no performance loss, which improves the power/thermal efficiency of the chip at no cost.

On-line characterization and reconfiguration for single event upset variations

“…Research into variation-aware adaptation is currently being conducted in multiple contexts, including timing [21] [14], leakage power [4], and temperature [15]. To [11].…”

Physical-aware task migration algorithm for dynamic thermal management of SMT multi-core processors

“…As mentioned earlier, the temperature of each core of a processor is different from other cores because of packaging technology [8], process variation [13] and the thermal effects of neighbor components [2]. Table I summarizes our experimental results of running different applications on four different cores of two Intel quad-core (Core i7-2600 and Core i7-3770) processors.…”

“…Different cores of a processor do not have similar thermal behavior due to process variation [13], the temperature effect of core neighbors [2], and other physical issues [1]. The temperature difference among cores of a processor running the same application can be as much as 10∼15…”

Physical-aware predictive dynamic thermal management of multi-core processors

“…The unique feature of the proposed algorithm is using what we call the core unique thermal behavior (CUTB) explained as follows. Different cores of a processor do not have similar thermal behavior due to process variation [18], the temperature effect of neighbor components [3], and other physical issues [1].…”

“…In response to these challenges, various Dynamic Thermal Management (DTM) techniques have been proposed to mitigate the thermal concerns of processors [2][3][4][5][6][7][8][9][10][15][16][17][18][20][21][22][23]. DTM is a set of techniques that control processor temperature at run-time so that temperature does not go beyond the critical temperature threshold.…”