In the ICT society, the number of Internet-connected devices and sensors continues to grow astronomically. Handling the huge amount of heat generated by data centers and IT devices is an urgent issue to ensure stable system operation as well as the health of the global environment.
Fujitsu has developed a novel liquid immersion cooling system that cools down servers by completely immersing them in liquid. This system won the 2018 Energy Conservation Grand Prize (for the first time) and the Environment Minister's Award for Global Warming Prevention Activity (second year in a row, sixth time overall). For this article, we interviewed Takashi Yamamoto, leader of the R&D project, about the process for commercializing this technology and its future potential.
Liquid Cooling: Tens to Hundreds of Times More Efficient Than Gas
Yamamoto: "The liquid immersion cooling system uses a technology that improves cooling efficiency by cooling down servers with liquid, though air is normally used. Compared to typical gasses such as air, liquid has a cooling capability that is tens to hundreds of times better. With a slow current, liquid can achieve significantly higher cooling performance than air without needing to maximize the current.
Devices are more stable when maintained at lower temperatures. Since liquid immersion cooling does not require the cooling fans and air conditioners needed for air cooling equipment, there are notable power savings as well."
Reviving Discontinued Technology to Meet Today’s Needs
"In fact, 30 years ago, a foreign company commercialized liquid immersion technology. I heard that it was considered to be a backbone technology for achieving super-high density packaging, which improved supercomputer performance back then. At that time, Fujitsu used water cooling technology (indirect liquid cooling in which a copper or other metal plate is put into water) rather than liquid immersion technology in order to cool down large computers. Thereafter, advances in semiconductor technology reduced power consumption significantly, and both liquid immersion and water cooling became unnecessary. When only air cooling technology came to be used, development of liquid immersion technology was suspended.
Due to its low viscosity, the coolant appears to be water at a glance. The liquid bath is needed to ensure a high degree of airtightness.
However, the amount of heat generated began to increase again as miniaturization and the speed of semiconductor devices advanced. Data centers now constitute more than 1% of global power consumption, and waste heat is also thought to be a problem due to heightened environmental awareness. Thus, a technology to replace air cooling is again needed.
Rather than cooling down specially designed supercomputers, we were motivated to reduce power consumption by effectively cooling existing general-purpose equipment that had heretofore been air cooled. So in a sense, we revived liquid immersion cooling technology."
Immersing in Liquid Requires Knowledge of Chemicals
"When considering a liquid immersion cooling system, the first key point is which type of liquid to use as the coolant. We chose a coolant based on various considerations: it does not conduct electricity; it is safe; it will not negatively affect server devices; it circulates smoothly because it is not so viscous; and so on.
When considering systems that customers will use for a long time, it is insufficient to merely cool the servers; it is also crucial to ensure ease of system operation and liquid bath airtightness to prevent coolant evaporation.
In designing conventional servers, knowledge of chemicals for choosing such a coolant or technology and a structural design to ensure airtightness are not necessary. But, because these servers must be immersed in liquid, we needed knowledge as well as new skills and know-how Fujitsu did not possess."
Completely Submerging Servers in Liquid to Cool Them Down with a "Liquid Immersion Cooling System" [FUJITSU JOURNAL] (Japanese)
Conventional Design and Development Processes No Longer Work
"Thus, we formed a team that includes members who studied chemicals at school back in the day. We also collaborated with Fujitsu Laboratories and our partners such as chemical suppliers to actively collect information and increase our knowledge.
To address structural issues such as airtightness, we created prototypes. In addition to using them for our experiments, we asked other departments as well as external customers and partners to use them to obtain know-how.
The standardized shapes of today’s servers and server racks have been in use for around 20 years. Their structure has not changed greatly, and there is a standard process for estimating the period and cost of new product design and development. To develop the liquid immersion system, however, that process did not work. Many unexpected events occurred and the process seldom advanced as we had envisioned. We faced the challenge of how to mature the process through repeated trial and error."
What was the key to realizing the liquid immersion cooling system? Let’s find out in Part 2 of this article.
An Episode from the Liquid Immersion Cooling System Lab:
A Wonderful Liquid, But Can You Drop a Smartphone in It?
As a proprietary technology, Yamamoto’s team obtained some patents for the liquid immersion cooling system that they developed. Among these patents are those for the liquid that maximizes cooling capacity, the fluorochemical coolant, and its working temperatures (the range of physical properties).
To determine the coolant liquid, the team focused on the following characteristics noted in the interview: no electrical conductivity, safe, no negative impact on servers, and especially a low viscosity. To circulate the flow so as to improve cooling efficiency, the liquid had to be less viscous. As a result, the team chose "3M™ Fluorinert™," which is a colorless, transparent, smooth, and odorless liquid that looks and feels exactly like water though it has totally different properties. It makes you feel as if you are submerging a machine in water. One day at the lab, a team member accidently dropped a smartphone into the liquid, but there was no need to worry. The smartphone worked fine without any problems.
Electric & Mechanical Engineering Development Divison
AI Platform Business Unit
After graduating from the Department of Mechanical Engineering of the Faculty of Engineering at Tokyo University of Science in 1989, Yamamoto joined Fujitsu Limited that same year and was assigned to the business division (Information System Business Division). Since then, Yamamoto has been involved in the development of cooling technologies and mechanical technologies for large-scale computer and storage products.