Thermal Management of Multi-Processors chips by Optimal Control Methods

نوع: Type: thesis

مقطع: Segment: masters

عنوان: Title: Thermal Management of Multi-Processors chips by Optimal Control Methods

ارائه دهنده: Provider: Porya Soltani Hanafi Por

اساتید راهنما: Supervisors: Abbas Ramezani (Ph. D)

اساتید مشاور: Advisory Professors: Mojtaba Hosseini Todashki (Ph. D)

اساتید ممتحن یا داور: Examining professors or referees: دکترا مجید غنی ئی زارچ دکتر زهرا سهرابی

زمان و تاریخ ارائه: Time and date of presentation: March,20,2021

مکان ارائه: Place of presentation: سمینار 2 برق

چکیده: Abstract: In today's processors, the integration of microelectronic circuits has reached the ultimate in compression. In this integration, to increase the computing power of the processor, the number of processor cores has been increased. Increasing the number of cores to a limited extent leads to an increase in power consumption density and an increase in temperature. Increasing the temperature reduces the reliability and life of the processor. Therefore, providing a solution for proper temperature and performance management in today's processors is inevitable. In this paper, a new algorithm for modeling and temperature management of multiprocessor systems using control theory methods is presented. since the processor being tested does not have a power sensor, we obtain the power consumption using the CPU Utilization method, and then using the obtained model, we obtain the thermal model of the processor by instantly recording the values of temperature and power. to obtain an accurate thermal model, we tested the processor with different benchmarks and performed sampling. The proposed algorithm controls the frequency of cores and fan speed online, depending on the load and temperature of the processor. The tests were performed on the ODROID-XU4 board, whose CPU consists of a Samsung Exynos 5422 multiprocessor chip. This board uses ARM big-LITTLE technology, which is a kind of architecture with asymmetric processors. This board has two quad-core processors. Unlike most work done, the effect of the controller on efficiency and temperature has been investigated simultaneously. The simulation results with the proposed algorithm show that by choosing the right cost function and its coefficients, a correct compromise can be made between CPU performance and temperature. Key Words: Multi processors, Simultaneous control of temperature and efficiency, Optimal Control, asymmetric processors

فایل: ّFile: Download فایل