Air cooling has long been the norm for datacenters. However, with next-generation chips exceeding 400W and calls for more conscious energy consumption, it’s safe to say that air cooling cannot accommodate this change and, subsequently, alternative methods are currently the hot topic of many conversations. Given that air cooling is ill-equipped to handle the considerable heat generated, liquid cooling is gathering traction for its distinctive approach of directly extracting heat from the source. Moreover, liquid cooling offers various advantages that can enhance overall data center operations.
If you’ve made it here, you’re probably curious about the differences between the liquid cooling techniques on offer. In this technology deep dive, we take an in-depth look into single-phase immersion cooling vs direct-to-chip cooling. What are the main differences between single-phase immersion cooling and direct-to-chip? Which one saves on your TCO and which one makes maintenance a breeze? Keep reading to find the right fit for you.
How Immersion Cooling and Direct to Chip Cooling Work
First things first, what does immersion cooling involve? What about direct-to-chip? Since both technologies are top contenders in the cooling market, let’s start with the basics of each one.
1. Single-Phase Immersion Cooling
In single-phase immersion cooling, the IT hardware is submerged in a specially formulated dielectric coolant that transfers the heat directly from the components. The immersion tank contains a CDU (coolant distribution unit) comprising a pump and a plate heat exchanger. The pump circulates the dielectric fluid within the tank, extracting heat from the servers. The heated dielectric fluid then passes through the plate heat exchanger, where it is cooled by water.
The coolant used in single-phase immersion is of course dielectric, meaning fluids based on hydrocarbon compounds such as vegetable oils, mineral oils, fluorocarbons, or other synthetic fluids (our SmartCoolant is a synthetic hydrocarbon).
Single-phase immersion doesn’t require airtight sealed vessels because hardly any coolant evaporates. Therefore, the vessels used in single-phase immersion are unsealed and known as “open baths”, where servers are placed side-by-side in large tanks that operate at atmospheric pressure.
2. Direct to Chip Cooling
Now that we know more about single-phase immersion cooling, let’s take a look at the basics of direct to chip cooling.
Direct to chip (DTC), a.k.a. cold plate cooling or direct liquid cooling (DLC), is a cooling method used for servers that involves circulating liquid directly over the components to dissipate heat. A flat, rectangular component with a chamber-like structure that is made of highly conductive material such as copper, known as a cold plate, is attached to the server’s heat-generating components.
The cold plate contains channels or tubes through which a coolant, typically water, flows. As the coolant passes through the chamber, it absorbs heat from the components, cooling them down. The heated coolant then exits the chamber and is either cooled in an external heat exchanger or expelled from the system. This direct contact between the liquid coolant and the server components allows for efficient heat transfer, ensuring effective cooling of the servers.
Since direct-to-chip dissipates heat directly from the hottest components e.g. the CPUs, this method typically uses air cooling in conjunction with the cold plates to cool the rest of the components that are not directly exposed to water cooling e.g. the DIMMS. Therefore, direct-to-chip cooling is the best choice for you if you’re dealing with high heat flux components i.e. the heat generated per unit of area.
Immersion Cooling vs Direct to Chip: Main Differences
What factors make single-phase immersion cooling better than cold plate or vice-versa? Let’s compare and contrast the main differences between single-phase and direct to chip.
- Thermal resistance: This is the ratio of the difference between the CPU temperature and incoming liquid temperature divided by watts. The greater the difference, the more heat is pulled away from the CPU. The lower the resistance, the better the cooling performance is at the CPU and heatsink level. At the moment, because direct to chip uses water, this more effectively transfers heat compared to immersion.
- TCO: The Total Cost of Ownership is an evaluation of the total cost of purchasing, deploying, and using and maintaining the product. For direct to chip cooling, the complete overhaul of servers tends to be exponentially more expensive. For immersion cooling, the initial cost is typically greater than that of air cooling but is quickly recompensated through energy savings.
- Sustainability: Both methods are a huge improvement from air cooling in terms of energy savings and a reduction in carbon emissions. Direct to chip cooling typically requires some air cooling. For this reason, immersion cooling has a lower PUE than that of direct to chip.
- Servicing & Maintenance: Servicing is cleaner with direct to chip cooling in the sense that the water flow can simply be stopped and the tubes can be unplugged, whereas immersion cooling requires interaction with the fluid.
- Risks: In the unlikely event of leakage, direct to chip cooling will lead to catastrophic damage of IT hardware and complete shutdown of the server.
- HW Flexibility: Immersion cooling is compatible with any IT hardware and retrofitting is a straightforward procedure. Adapting traditional servers for direct to chip cooling requires a much more intricate and complex process.
- Heat reuse: Both methods provide the potential for heat reuse and its subsequent opportunity for another source of revenue for data centres. Due to its full coverage, immersion cooling can capture all heat across all components. While direct to chip has less coverage, it can obtain higher temperatures. Applications for heat reuse include district heating, radiant floor heating systems, and hot water uses.
Why We Choose Single-Phase Immersion
While the benefits of direct-to-chip cooling are clear, immersion cooling has consistently remained the top choice for us at Submer. It earns its place as our favorite through compatibility with all IT hardware, ability to capture 100% of the heat generated, and its safety for operators and environmental friendliness. Immersion cooling really strikes the ideal balance between performance and sustainability.
Whichever liquid cooling technique you choose to implement, you’ll not only be staying apace with high-performance computing and future-proofing your facilities, but also paving the way forward for conscious energy consumption in datacenters. Who knows what innovations lie ahead for the world of liquid cooling? Watch this space…
Still not 100% convinced which technology is for you?