Home / Guides / Data centers

Liquid-to-liquid or liquid-to-air CDU: compare the interfaces

Compare CDU architectures by heat-rejection path, facility services, space, and operational responsibilities.

US Liquid Cooling editorial · Reviewed · 4 min read

Choosing between liquid-to-liquid and liquid-to-air cooling distribution begins with a facility question: where will the collected heat go? Both approaches can support liquid at the technology side, but they hand the heat to different surroundings. The correct comparison considers the complete thermal path rather than the CDU label alone.

This article provides a procurement framework. The final selection and operating limits require the server vendor, cooling supplier, and qualified facility-design team to work from the same requirements.

Draw both proposed heat paths

For a liquid-to-liquid option, identify the technology circuit, the heat exchanger, and the facility liquid circuit receiving the heat. Then continue the drawing to the equipment that rejects or reuses it. Include any intermediate equipment that is essential to operation.

For a liquid-to-air option, identify the air stream receiving the heat and how that heat is subsequently removed from the space. Avoid interpreting the lack of a facility-water connection as the absence of a facility-cooling responsibility. The surrounding environment is still part of the design.

Establish the available site services

Document whether suitable facility water is available at the planned location. If it is, obtain its actual design and operating conditions rather than assuming an existing pipe is suitable. If it is not, establish what modification would be required and who would approve it.

Record electrical requirements, floor or rack space, access routes, and service clearance for each alternative. A solution that fits in a product photograph may be difficult to install in a populated room. Include the path through doors and aisles in the practical review.

Compare on common operating assumptions

Give both suppliers the same server load assumptions and environmental requirements. Ask them to identify where their proposed performance depends on conditions outside the CDU itself. This makes the quotations easier to compare and exposes differences that a headline rating can hide.

Request performance information across the expected operating range, not only a single nominal point. Have the engineer explain how room conditions, facility conditions, or seasonal operation influence the design. Do not convert an example from another installation into a guaranteed result for your site.

Include the residual cooling requirement

Ask the server manufacturer which components remain air cooled. Coordinate that requirement with the selected liquid-cooling architecture. The room’s existing air-cooling system may continue to serve a necessary purpose even after the main processors receive liquid cooling.

A useful design review traces every significant heat source to its destination. It also identifies which monitoring system confirms acceptable operation. This prevents the project from treating the liquid-cooled portion as the only part of the server environment that matters.

Evaluate service and expansion

For each option, describe the maintenance arrangement, expected isolation boundaries, and access to replaceable components. Ask how a future rack is added and which infrastructure would need modification. Expansion should be represented in a coordinated plan, not only in spare capacity on one component.

Consider organizational ownership as well. A deployment may span IT operations, facilities, and an external service provider. Identify who controls setpoints, maintains the heat-rejection equipment, and responds to an alarm that crosses those boundaries.

Build a comparison that supports a decision

Use columns for server compatibility, required facility services, installation work, environmental assumptions, maintenance access, controls integration, and future expansion. Keep cost items associated with the party supplying them so exclusions remain visible.

Vertiv’s product portfolio illustrates that liquid-cooling distribution can be offered with several heat-transfer arrangements. It does not establish that one arrangement is universally better. Choose the architecture whose complete operating assumptions match the facility, and retain those assumptions in the project record.

Sources & further reading

Use the manufacturer’s instructions for your exact equipment. Read our editorial policy or suggest a factual correction.

Explore the company directory → · All liquid-cooling guides →

01
Data centers

Direct-to-chip liquid cooling: following the heat path

Understand cold plates, server connections, distribution, and heat rejection before comparing direct-to-chip systems.

↗
02
Data centers

CDU Approach Temperature: Read the Capacity Rating

Learn what CDU approach temperature means, separate it from loop temperature rise, and compare capacity quotes using the same fluids and design conditions.

↗
03
Data centers

Connect liquid cooling to a heat-rejection plan

Understand why the outdoor heat-rejection system, operating conditions, and site constraints matter as much as the server cooling loop.

↗

Directory matches are starting points for research. Confirm product fit and availability with the company.