Here is the abstract you requested from the Thermal_2016 technical program page. This is the original abstract submitted by the author. Any changes to the technical content of the final manuscript published by IMAPS or the presentation that is given during the event is done by the author, not IMAPS.
|Optimization of Heat Sinks, Server Orientations and Channels to Decrease the Junction Temperature for Oil-Immersed Servers|
|Keywords: Oil Immersion Cooling, Shadow Effect, Ducting over heat sinks|
|With the aggressive implementation of Internet of Things and Cloud Computing, hardware developers and cooling solution providers face a constant challenge. To reduce the PUE’s and thermal resistance across high end servers datacenter cooling strategies have changed over time. One of the alternative is to replace forced convective air cooling method at the server level with a forced convective low velocity non-conductive mineral oil cooling that has high heat carrying capacity. Oil immersion seems to be a promising alternative to mitigate the challenge of increasing junction temperatures when the heat flux density increases exponentially. This is required even more when one heat generating element is placed in the shadow region of another to reduce the server’s foot print. In this study, Open Compute Server Generation - 3 that has a similar configuration is considered. The baseline model has air pumping volumes of chilled air from one end of the server. This model is compared with mineral oil cooled system and subsequently, is compared with the geometrically optimized model for oil cooling. These there models are compared at constant mass flow rate and constant pumping power. As the oil cooling system does not require fan for displacing the media the fan components are removed and the height of server is reduced considerably. This helps in stacking more servers within the same space. The heat sinks geometries on both the chips are optimized to reduce the overall junction temperature. Ducting is incorporated over the RAM’s to guide majority of the flow over the heat sinks. Moreover, different server orientations are studied to see the effect of gravity on thermal shadowing, junction temperatures and pumping power. All the Numerical Analysis are conducted using ANSYS Icepak software. The optimizations are conducted by using parameterization tools. The resultant effect of using mineral oil and these optimizations is that the high-end servers can have much higher heat flux density nevertheless maintain its junction temperature under acceptable limits. Electronics, MEMS and Nano Electronic Systems Packaging Center, University of Texas at Arlington Principal Investigator (firstname.lastname@example.org)|
|Kunal Atulkumar Shah, Graduate Student
University of Texas at Arlington