Here is the abstract you requested from the Thermal_2014 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.
|Making device level liquid cooling in the real world|
|Keywords: Phase change, Device-level cooling, Data center thermal management|
|Device level liquid and two-phase cooling technologies have existed for decades, but it is only relatively recently that practical implementations have shown up outside of HPC or high density power electronics applications. Why is this? This presentation will present a brief historical perspective on liquid cooling technologies, ending with a two-phase pumped refrigerant implementation as a case study. Although nearly 12 years of laboratory research were expended with the focus on heat removal at the device, it seems that this may have been the easy part. Getting the technology out of the laboratory and into an environment like a datacenter requires a system that is designed with the physics of heat removal always in sight. First, the thermodynamics of two-phase heat transfer, while always obeying the laws in the classic engineering textbooks, are often far subtler than a textbook treatment. The mini-evaporator that worked great in the lab often behaves very differently in the field because of pressure fluctuations and shifts in saturation temperature driven by the heat load. We will show how careful design is needed to prevent heat-load driven flow variations and how saturation temperatures throughout the system can be made much less heat-load dependent. In addition, pumped systems with phase change are particularly susceptible to instabilities caused by two-phase flow behavior and pressure fluctuations at the device, the rack and the system levels. Additional design strategies are presented that have been effective in the field that prevent boiling instabilities from transmitting backward or forward through the flow network, that allow multiple multiphase fluid paths to be connected in parallel with minimal cross-talk, that allow hot-swapping of fluid paths during operation without affecting other fluid paths and that prevent large scale two-phase flow instabilities in the piping network that distributes the fluid to multiple racks. This combination of design strategies results in a scalable, effective pumped refrigerant implementation that can remove 2000 W or more from each server in a row of densely packed racks.|
|Timothy A. Shedd,