Thermal study of a Raspberry Pi for the construction of a small-sized supercomputer
This project is part of a series of projects whose final purpose is to design a very small computing cluster coupling up to 100 units of Raspberry Pi 2, equivalent to a system with 400 CPUs. The main advantage of a cluster like this it is its reduced size and a substantial cost-cutting on the invest...
Gespeichert in:
1. Verfasser: | |
---|---|
Format: | Dissertation |
Sprache: | cat |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | This project is part of a series of projects whose final purpose is to design a very small computing cluster coupling up to 100 units of Raspberry Pi 2, equivalent to a system with 400 CPUs. The main advantage of a cluster like this it is its reduced size and a substantial cost-cutting on the investment. As one of the two first projects of these series, the present one aims to thermally characterize the Raspberry Pi 2 and propose a feasible cooling technology. The most
critical issue of such a cluster is the cooling design in order to ensure a proper CPU temperature. Indeed, if the temperature exceeds a threshold of 85ºC, the device can crash [1]. However, such temperature is too high to guarantee an optimal performance of the Raspberry. To this aim, our goal is to maintain the device temperature below
60ºC under extreme atmosphere conditions, i.e. at an air temperature of 40 ºC.
The first task within present study has been to experimentally detect and quantify hot spots in the Raspberry Pi 2, when working at full capabilities. This information is required to decide which points need a special cooling technology. In order to reduce the experimental costs, the viability of using a NTC thermistor and a microcontroller
has been studied and compared with the classical type-K thermocouples connected with the data logger. To this aim, a preliminary experiment has been designed to calibrate a NTC thermistor The second task has been to calculate the thermal power generated in the different components of the device. To this purpose, the data reported in the experiment for the detection of hot spots has been used. Also, the thermal convection coefficients required to calculate the thermal power generated have been estimated using available correlations in the literature. The third and last task of the project has been to determine experimentally if the available heat transfer coefficients in the literature are valid in the presence of fins designed for small devices as the Raspberry Pi 2. This task includes two different experiments: one for natural convection and another one for forced convection using a small wind tunnel.
Some Arduino and Raspberry PI knowledge is required to carry out the three tasks. In the document, the basic information has been provided so the reader can manage to keep on the reading. Moreover the sketches are completely commented and several tutorials are attached to help the reader understand the content of this document easily. With the su |
---|