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.
Gravity-driven cooling systems employing microchannel flow boiling can become a scalable and viable long term solution for the cooling of datacenter servers. In order to design this emerging type of cooling system, a new dynamic simulation tool using interconnected PDEs is described in the first part of the present paper followed by the validation of the modeling for both steady and dynamic regimes using a first-of-a-kind 15 cm-height thermosyphon test bench. A corrected error of 0.4 K was found for 7 different steady-state mean chip temperatures and heat load disturbances were predicted with the same 2-stage process presented here as the experimental results. The code was then used to predict the behavior of a mini-thermosyphon that would fit within the height of a 2U server. Multiple steady-state solutions were found with both stable and unstable operating states. A sensitivity study demonstrated that a 30% increase in the riser internal diameter led to 10-60% increase in the mass flow rate depending on the heat flux. Then, simulations with unbalanced heat loads showed that the flow gets self redistributed into the CPU with the highest heat load, e.g. 1.7 times larger flow rate for a heat flux 3 times larger. Finally, heat load and water coolant flow rate disturbances were also simulated and discussed. (C) 2016 Elsevier Ltd. All rights reserved.
Remco Franciscus Peter van Erp
, , ,
John Richard Thome, Jackson Braz Marcinichen, Raffaele Luca Amalfi, Filippo Cataldo