Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
The main objective of this paper is to utilize an improved version of the simulator presented at InterPACK 2017 to design a thermosyphon system for energy-efficient heat removal from 2-U servers used in high-power datacenters. Currently, between 25% and 45% of the total energy consumption of a datacenter (this number does not include the energy required to drive the fans at the server-level) is dedicated to cooling, and with a predicted annual growth rate of about 15% (or higher) coupled with the plan of building numerous new datacenters to handle the "big data" storage and processing demands of emerging 5G networks, artificial intelligence, electrical vehicles, etc., the development of novel, high efficiency cooling technologies becomes extremely important for curbing the use of energy in datacenters. Notably, going from air cooling to two-phase cooling, not only enables the possibility to handle the ever higher heat fluxes and heat loads of new servers, but it also provides an energyefficient solution to be implemented for all servers of a datacenter to reduce the total energy consumption of the entire cooling system. In that light, a pseudo -chip with a footprint area of 4 x 4 cm2 and a maximum power dissipation of 300 W (corresponding heat flux of about 19 W/cm2), will be assumed as a target design for our novel thermosyphon-based cooling system. The simulator will be first validated against an independent database and then used to find the optimal design of the chip's thermosyphon. The results demonstrate the capability of this simulator to model all of the thermosyphon's components (evaporator, condenser, riser and downcomer) together with overall thermal performance and creation of operational maps. Additionally, the simulator is used here to design two types of passive two-phase systems, an air- and a liquid-cooled thermosyphon, which will be compared in terms of thermal -hydraulic performance. Finally, the simulator will be used to perform a sensitivity analysis on the secondary coolant side conditions (inlet temperature and mass flow rate) to evaluate their effect on the system performance.
Remco Franciscus Peter van Erp
David Andrew Barry, Ulrich Lemmin, Naifu Peng, François Mettra, Rafael Sebastian Reiss