Here is the abstract you requested from the Thermal_2017 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.
|Passive Two-Phase Immersion Cooling Extended Run Test Vehicle Results and Observations|
|Keywords: Immersion, Extended, Data|
|High Performance Computing (HPC) systems continue to push the limits of current thermal management methods. Conventional air cooling techniques are often unable to extract sufficient heat, requiring advanced solutions such as liquid cold-plate cooling. An alternative thermal management approach is passive two-phase immersion cooling, where electrical components are submerged into dielectric fluid baths. Previous work presented at this conference in 2010 and 2012 demonstrated efficient heat removal from electronics with this passive two-phase immersion cooling technique using Novec 649 fluoroketone dielectric fluids from 3M Company 1,4. Experimental results from our previous work indicated that high- speed links, which operate above 15 GHz bandwidth, as well as co-planer optical interconnects performed on par with air cooled equivalent designs. In addition, the uniform thermal environment provided by passive heat transfer technology does not require the use of pumps, fans, cold plates, or manifolds, thereby enabling increased packaging densities for HPC system designs. However, passive two- phase immersion cooling using fluoroketone dielectric fluid is not without its own set of challenges. In particular, fluid loss mechanisms, material compatibility and system design considerations must be taken into account to make two-phase immersion cooling a viable thermal management technology 2,3. Mayo Clinic has been conducting an extended run passive two-phase immersion cooling experiment using a Dell Optiplex 755 PC equipped with a solid state hard drive 4. While there are many examples of short-term experiments documenting the heat transfer performance of passive two-phase immersion cooling at the component and system level, the primary focus of this presentation will be the heat transfer performance characteristics of this technology over a prolonged period of time (beyond 5 years) using Novec 649 dielectric fluid 1-5. The test system has been running under approximately 100% computational workload using Prime95 while immersed in Novec 649 for more than 6.5 years without a single electrical component failure. During the extended test the processor load, temperature, and dielectric fluid performance property datum have been logged and collected, the results of which will be presented. Time lapse images collected from inception to present day will also be presented documenting physical changes during the “aging process” of this particular test platform. Our talk will conclude with a summary of system level design considerations and requirements learned from the Mayo Clinic test platform.|
|Nicholas A. Klitzke, Mechanical Engineer