Indirect on-chip power and temperature distribution sensing for hotspot-limited microprocessors

Abstract

Recent trend in microprocessors gave rise to needs for a dense thermal/power management. However, due to the highly limited area budget of silicon, the development of dense sensing techniques has been faced with difficulties. This motivated us to aim for a radical change in sensing methodology. In this thesis, we propose a set of techniques for an area-efficient dense thermal/power map acquisition. By regarding the thermal/power map as spatial information, i.e. image, we have successfully introduced image processing and computer vision techniques into thermal map sensing field. First, we propose a new temperature-to-power technique, named PowerField, supporting both transient and steady-state analysis based on a probabilistic approach. Unlike the previous works, PowerField uses two consecutive thermal images to find the most feasible power distribution that causes the change between the two input images. To obtain the power map with the highest probability, we adopted Maximum-a-posteriori Markov random field (MAP-MRF). For MAP-MRF framework, we modeled the spatial thermal system as a set of thermal nodes and derived an approximated transient heat transfer equation which requires only the local information of each thermal node. Experimental results with a thermal simulator show that PowerField outperforms the previous method in transient analysis reducing the error by half on average. We also show that our framework works well for steady-state analysis by using two identical steady-state thermal maps as inputs. Experimental results determining the binary power patterns of an FPGA device is presented achieving 90.7% average accuracy. The second half of this research presents an area-efficient thermal sensing technique: hybrid temperature sensor network. The proposed sensor architecture fully exploits the spatial low-pass filtering effect of thermal system, which implies that most of the thermal information resides in very low spatial frequency region. Our on-chi...