There was a time when increased computing power was a matter of space. The only way to keep servers running at peak capacity was to blow cool air on them. Today, many organizations' environments operate effectively under this and other methods we review below. However, immersion cooling is often the method of choice for companies that need increased performance and efficiency.
Again, the server rooms of many companies are functioning under this traditional cooling method. An entire market of CPU fans exists to aid the performance of conventional and entry-level servers. These usually smaller companies can operate efficiently with the equipment they have.
At the same time, CPUs as recently as five or six years ago operated primarily in a power range of 125 to 150 watts. In contrast, many of today's high-performance CPUs require 350 to 400 watts. Therefore, even though computing power continues to increase, there's a price for data center operators to pay in the form of heat rejection.
Direct-To-Chip Liquid Cooling (DCLC)
As supercomputers developed to serve academic and government organizations, their specialized hardware faced cooling dilemmas. One initial approach was to target the hottest components (i.e., CPUs) and deliver liquid-based cooling directly to them, which is DCLC.
Today, many specialized data centers perform effectively using this strategy. However, the design of a data center and corresponding rack system must accommodate its use. And, making the approach more complicated, DCLC cooling technology is hardware-specific, making switching vendors and upgrading within a vendor's hardware challenging.
For most organizations' growing cooling needs, immersion is the flexible and scalable choice. It entails the submersion of an entire server in a fluid (usually an oil) that acts quickly to remove heat, similar to how a blacksmith might cool a piece of molten metal in water. In addition, the fluid acts as a better conductor of heat than air, so less is needed, and components can exist in a smaller space.
There are multiple advantages of an immersion approach over air-based cooling. The first is the ability to condense more computing power into a single rack or racks within a facility. Air-cooled servers require more room so that air can flow through CPUs causing fewer CPUs to fit on each rack. And with fewer servers per rack, the entire facility needs to grow to accommodate more computing power.
Because of the effectiveness of immersion cooling, CPUs and GPUs can operate closer together, thereby using less rack space. For example, an air-cooled rack system requiring 8U may fit in an immersion-cooled rack of 1U or 2U in size. Therefore the overall density of computing power within a given space is higher.
Immersion Cooling Use Case #1: Peak Performance
Companies working in the HPC or financial space often need their computing platforms to run faster. For them, time is money. Immersion cooling enables data center designers in these markets to pack more servers in less space and run their CPUs at overclocked speeds.
Immersion Cooling Use Case #2: Reduced Energy Cost
In areas where energy costs are high or local governments enforce efficiency standards, immersion cooling enables data centers to operate at optimal ratios. That means their energy expenditure to cool their computing environments is low compared to the servers' power. Ideally, a newer data center should target a 1.1 ratio meaning its cooling power consumption is only 10% of its overall computing consumption.
Immersion Cooling Use Case #3: Scalability
Finally, whether a company targets maximum performance or power efficiency, the ability to grow is often a concern. Unlike in the case of air-cooled environments, liquid-cooled technology allows for the swapping out of CPUs and GPUs without the need for more space or upgraded fans. And similarly, compared to DCLC, immersion technology isn't hardware dependent. Therefore a change in processor or other components doesn't require a modification in cooling hardware.
Single vs. Two-Phase Immersion Cooling
Single-phase immersion cooling involves using a heat exchanger to remove the heat from the fluid without boiling. The fluid flow through the system is the primary way heat transfers through the system. This method is more straightforward and less expensive of the two options and is used in most immersion cooling operations.
Two-phase immersion cooling involves boiling liquid to transfer heat instead of flowing to a heat exchanger. Then, a condenser recaptures the fluid for re-use in the system. While this option requires a higher cost of entry, it offers the potential for the lowest cost of ownership over time.
Designing Your Next Cooling Solution
As you can see, immersion cooling offers data center designers the opportunity to increase their systems' performance while maintaining the most efficient operations.
So, when it comes time to design your next cooling solution, look to UNICOM Engineering to provide the proper hardware and expertise to support your latest data center design. In addition, our experts will ensure you have the computing power you require today in a design that scales to meet the needs of tomorrow.
As an Intel Technology Provider and Dell Technologies OEM Partner, we can supply, build, and support the best hardware to meet or exceed the needs of your application and help you bring it to market faster. Schedule a consultation today to learn more about how UNICOM Engineering can keep you moving forward.