Paper

Two-Phase Cooling System Performance Under Different Operating Scenario

Abstract

Performance needs of today’s AI systems is driving increase in power density of high-performance processing units and integration into dense 2.5D and 3D heterogeneously integrated packages that exceeded the capability of traditional energy-intensive air-cooling solutions used in Data Centers. There is a need to develop both single- and two-phase liquid cooling solutions to provide energy-efficient thermal management of high-performance AI systems while eliminating water consumption and maintaining system reliability in any geographical location. In this work, previously developed component-/system-level models [1] and Sim2Cool tool [2] were adapted to evaluate the performance of above-ambient pumped two-phase liquid cooling of a multi-server rack under different operating scenarios and geographical locations. The tool was used to compare both component level as well as system level performance impact (in terms of thermal resistances, pumping power, cooling coefficient of performance, junction temperatures, etc.) of utilizing different coolants (such as R1234ze(E), R1234ze(Z), R1233zd(E), etc.) in the two-phase cooling loop. The system level model was utilized to understand the two-phase cooling system performance under different operating scenarios including (i) over a typical meteorological year at different US cities, (ii) for different facility side temperature set-points, (iii) for different internal/external loop pump and dry-cooler fan speeds, (iv) for different refrigerants at the same flow rate or at the same target exit vapor quality, etc. Results indicate medium-pressure (e.g. R1234ze(Z)) or high-pressure (e.g. R1234ze(E)) refrigerants show a better component as well as system level performance compared to low-pressure (e.g. R1336mzz(Z)) refrigerants.

Acknowledgment: “The information, data, or work presented herein was funded in part by the Advanced Research Projects Agency-Energy (ARPA-E), U.S. Department of Energy, under Award Number DE-AR0001577. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.”

References:

  1. Parida, P. R., Chainer, T., Iyengar, M., David, M., Schultz, M., Gaynes, M., Kamath, V., Kochuparambil, B., Simons, R. and Schmidt, R., System-Level Design for Liquid Cooled Chiller-Less Data Center, ASME IMECE 2012, Houston, TX, November 2012.
  2. P. R. Parida, M. D. Schultz and T. Chainer, "Sim2Cool: A Two-Phase Cooling System Simulator and Design Tool," 2018 17th IEEE Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems (ITherm), San Diego, CA, USA, 2018, pp. 271-280, doi: 10.1109/ITHERM.2018.8419572.