Rear doors are usually bought to solve a density problem inside an existing hall, so the supplier's ability to survey, install, and service in place matters as much as the coil.
Infrastructure vendors
Vertiv, Schneider Electric, nVent, STULZ, and Rittal supply rear-door units alongside racks, CDUs, and plant equipment, which keeps the hydraulic boundary and the service contract with one party.
Liquid-cooling specialists with plant depth
Motivair pairs rear-door units with CDUs and chillers, which suits a project that needs the whole chain from rack to heat rejection rather than a coil alone.
What to specify
Rear doors are rated at a water temperature and an air-side condition, and both change once the hall is loaded. Ask for capacity at your actual supply temperature and return air, with the fan and control strategy that achieves it.
- Capacity at site supply water and return air temperatures
- Fan redundancy and behavior on power loss
- Condensation strategy and leak detection
- Rack compatibility and door swing clearances
Why RDHx units suit high-density retrofits
A rear door heat exchanger removes heat from the rack exhaust before it mixes into the room, which is what allows an existing data center to raise rack density without rebuilding its air path. Because an RDHx is indifferent to what is installed in the cabinet, it avoids the hardware qualification that direct-to-chip liquid cooling requires, and it can be redeployed when the fleet refreshes. Chilled water or facility water has to reach the rack, and that pipework, not the coil, is usually the constraint on how many rows can be converted.
Active and passive units behave differently
Passive units rely on server fans to push air through the coil, which keeps the cooling system simple and adds no fan energy but limits capacity and depends on the servers' own airflow. Active units add fans to the door, raising capacity and giving control over the air-side condition at the cost of energy efficiency and another component to power and maintain. High-performance deployments generally need active units; the specification should state which is offered and what happens to the rack on loss of door power.
Scaling RDHx solutions across an HPC hall
One rear door proves the concept; a high-performance computing hall needs the approach to be scalable across rows. That means water reaching every cabinet that may need it, a control strategy that handles rows at different densities, and enough spare pumping and plant capacity for the rows still to be converted. Because the units mount to standard 19-inch racks and leave the servers untouched, thermal management can be raised row by row as demand arrives, which is the main reason this equipment survives fleet refreshes better than in-server cooling does.
Where air cooling still carries the load
A hall using rear-door units is still an air-cooled data center in the sense that servers move air across their own components; the doors simply intercept the heat earlier. Room air handling therefore remains part of the cooling system and has to be sized for the residual load, for cabinets without doors fitted, and for the condition of the hall if door fans are lost. Treat the two as one design rather than assuming the coils displace the room entirely.
Condensation is the failure mode to design out
An RDHx carries water into the hall and cools air below room temperature, so if supply water drops below the dew point the unit will condense. The usual protection is to run the loop warm, above dew point, and to monitor it continuously rather than to rely on drainage. Ask how the controls enforce the dew point limit, what the leak detection covers, and how a unit is isolated without taking the rack offline.