Will RDHx improve data center efficiency and PUE?

Rear-door heat exchangers (RDHx) can improve data center efficiency and reduce PUE—but the size of the improvement depends on what you change (fans, chillers, pumps), where you measure (rack vs plant), and how you operate the water temperatures and economizers.

This guide is written for facilities and operations teams. It focuses on the levers you can actually pull (and verify) at the site level: fan power, pumping power, chilled-water return temperature, plant ΔT, and economizer hours.

Key Takeaway: RDHx often reduces air-side cooling work, but the biggest site-level gains typically come when RDHx enables warmer water and more economizer/free cooling hours—without creating hidden pump, controls, or condensation penalties.

Table of Contents

Start with the PUE math (so “10% savings” doesn’t mislead you)

PUE = Total Facility Energy / IT Energy. A change in cooling energy affects the numerator. A change in server fan energy affects the denominator too, because server fans are part of IT power.

For a simple baseline, assume:

IT load = 1.0 MW
Baseline PUE = 1.4 → total facility power = 1.4 MW
Non-IT overhead = 0.4 MW

If (and this is a simplifying assumption) that 0.4 MW overhead is entirely “cooling + fans,” then:

10% reduction in overhead → PUE ≈ 1.36
20% reduction → PUE ≈ 1.32
30% reduction → PUE ≈ 1.28

That’s why you can hear “we cut cooling energy by 20%” and still see a modest absolute PUE change. (And if RDHx increases pumping power, the net improvement can be smaller than expected.)

For a canonical definition of PUE, see TechTarget’s overview of power usage effectiveness (PUE).

Best practice 1 — Separate “air-side fan savings” from “water-side pump overhead”

RDHx shifts a portion of heat removal from room air cooling to rack-level heat-to-water capture. That can reduce the burden on CRAHs/CRACs and in-row units, but it also introduces water-side power.

Where fan savings typically show up

Facilities teams most often see savings (or avoided growth) in:

CRAH/CRAC fan power (lower airflow demand, less bypass/recirculation)
in-row fan power (where in-row is used to manage hot spots)
sometimes server fan power (a real fan power reduction can occur if inlet conditions stabilize and the server fan curve backs off)

Whether server fan power goes down depends on inlet temperature stability, pressure drop, and airflow paths. Don’t assume it—meter or infer it from IT power breakdown if available.

Where pump power shows up (and why it matters)

Most RDHx implementations require a water loop with:

flow through the rear-door coil
valves and controls
potentially a secondary loop or a coolant distribution unit (CDU) depending on architecture

Pumping energy is the most common “silent offset” to air-side savings. If your loop is high head (long runs, restrictive passages, poor balancing), pump kW can rise enough to dilute the PUE win.

Failure mode: savings erased by high head / over-pumping

If you’re chasing a PUE improvement, treat “design and operate to the minimum required flow” as a first-class objective:

right-size the control valve and verify stable modulation
avoid bypass/mixing that destroys ΔT
verify pressure drop and pump curve against actual operating points

Best practice 2 — Use RDHx to improve plant ΔT and return temperature (not just rack temps)

Many teams evaluate RDHx as a “rack temperature fix.” That’s necessary—but not sufficient if your goal is efficiency and PUE.

A more profitable lens is: Does RDHx improve chilled-water return temperature and plant ΔT (delta-T)? If it does, you can reduce required flow and improve chiller and pump efficiency.

Why plant ΔT matters

Plant ΔT is the temperature rise between chilled-water supply and return. Low ΔT forces more flow to move the same heat, which increases pump energy and can cause inefficient chiller staging.

For a clear explanation of why low ΔT wastes energy and increases flow/pump cost, see How to get the most from your chilled water plant.

How RDHx can raise return temps (and support warmer water)

DOE/LBNL’s guidance notes RDHx can perform well at warmer chilled-water setpoints, and in some configurations can reduce or even avoid chiller energy use by enabling alternative heat rejection paths when conditions allow (for example, a plate-and-frame heat exchanger with tower operation). DOE/LBNL documents this in “Data Center Rack Cooling with Rear-door Heat Exchanger.”

This is where plant operators often see the real leverage:

higher supply temperatures reduce compressor lift
higher return temperatures help maintain design ΔT
higher temperature tolerance can expand economizer hours

Failure mode: bypass/mixing and low-ΔT syndrome persists

RDHx won’t automatically “fix” low ΔT if the rest of your system mixes return streams or over-controls coils.

What to watch:

return temperature trends versus load
valves pinned open with low ΔT (a sign of control or sizing issues)
bypass lines that keep return water artificially cool

Best practice 3 — Treat economizers/free cooling as the main prize (when climate + temps allow)

RDHx can improve PUE by lowering air-side fan energy, but many sites see the biggest step-change when RDHx helps them operate at warmer water temperatures and run economizers longer.

What has to be true for more economizer hours

Facilities teams should validate these prerequisites early:

allowable IT inlet temperature envelope and hot-spot tolerance
controls sequence that actually switches to free cooling (including a waterside economizer) when conditions permit
sufficient heat exchanger approach (tower + plate-and-frame sizing)
water quality program suitable for the chosen heat rejection method

Failure mode: setpoints too conservative; economizer sequence not used

Two common patterns kill economizer benefits:

Operators keep water setpoints low “just in case,” eliminating free cooling hours.
Economizer sequences exist on paper but are not used due to nuisance alarms or unstable control.

If your economizer strategy is the main reason you expect a PUE change, test it during commissioning—not months later.

Best practice 4 — Commission to dew point, leaks, and controls (the hidden PUE killers)

RDHx introduces new operational constraints. The PUE story doesn’t hold if the system is unstable, alarm-prone, or risky enough that operators disable it.

Dew point margin and condensation controls

Condensation risk is real if coolant temperatures go below room dew point. DOE/LBNL commissioning guidance explicitly calls out maintaining coolant inlet temperature above dew point.

⚠️ Warning: If you can’t maintain dew-point margin seasonally (humidity swings, outside air events, control drift), your “efficiency” project can become a reliability incident.

Leak detection, water quality, and maintainability

Your risk controls should be explicit:

leak detection at connections
isolation strategy (valves, quick disconnects where appropriate)
water quality management (corrosion/fouling prevention)
maintenance access (rear clearance, door servicing)

Failure mode: alarms, instability, and “operators disable it”

A technically efficient system that is operationally fragile often gets set back to conservative defaults. In practice, this is one of the most common reasons expected savings never show up in the monthly bill.

Best practice 5 — Validate with a measurement plan (before/after and normalized)

If you want an honest answer to “Will RDHx improve PUE?”, you need a measurement plan that survives real operating variability.

DOE/LBNL’s energy-efficiency action lists emphasize sensor placement that reflects IT intake conditions, calibration, and metrics used to assess airflow effectiveness.DOE/LBNL “Data Center Master List of Energy Efficiency Actions”

Minimum metering points (practical list)

At minimum, plan to trend:

IT kW (or PDU output) and total facility kW (for PUE)
CRAH/CRAC fan kW (or VFD speed + inferred kW)
RDHx loop pump kW (and any CDU pump kW)
chilled-water supply temperature, return temperature, and flow (to compute ΔT and heat moved)
economizer status and hours (waterside and/or airside)
room dew point (or T/RH from which dew point is computed)

The plots to review monthly

Build a simple monthly review pack:

total facility kW vs IT kW (trendline)
cooling system kW (fans + pumps + chillers) vs IT kW
chilled-water ΔT over time (and during peak load periods)
economizer hours by month versus outdoor conditions
alarms/events correlated with setpoint changes

This normalizes for workload changes and avoids the most common trap: comparing two different months with different IT load and weather and calling it “savings.”

What PUE improvement is realistic?

A realistic answer depends on your baseline and what portion of overhead you can change.

Directional guidance:

If baseline PUE is already low (e.g., near 1.2), RDHx may still help with hotspots and density, but the absolute PUE drop may be small because the overhead slice is already thin.
If baseline PUE is higher (e.g., 1.5–1.8) and a meaningful portion is driven by air movement and compressor hours, RDHx paired with warmer-water operation and economizers can produce more visible site-level improvements.

The right way to state the expectation is:

estimate the fraction of non-IT overhead you expect to reduce (air-side)
subtract the expected increase (or decrease) in pump and door-fan power
model the plant benefit if warmer water increases economizer hours or reduces lift

For a discussion of RDHx optimization and net energy effects under modeled conditions, see the Copper Development Association’s article on optimized rear-door heat exchangers (RDHx).

FAQ

Does a rear door heat exchanger PUE improvement always happen?

No. RDHx often reduces room-level cooling burden, but whole-site PUE improves only if the net effect reduces total facility energy. Added pumping power, conservative setpoints, or limited economizer use can flatten the PUE change.

Can RDHx increase pump energy?

Yes. RDHx introduces water-side flow and pressure drop. If the loop is high head or over-pumped, pump kW can increase and offset air-side fan savings.

Does RDHx work with warmer water?

Often, yes—within your IT inlet limits and dew point constraints. DOE/LBNL notes RDHx can perform well at warmer chilled-water setpoints, which is part of why it can reduce plant energy under the right conditions.

How does RDHx compare to room-level air cooling for efficiency?

RDHx reduces hot exhaust recirculation into the room and moves heat to water at the rack. That can reduce CRAH/CRAC burden and can improve plant operation (ΔT and chiller lift) when setpoints are adjusted accordingly.

For a related deep-dive on RDHx behavior and capacity concepts, see Coolnetpower’s guide on how much heat a rear door heat exchanger can remove. For a structured measurement approach to liquid cooling pilots, see how to measure PUE in a liquid-cooling PoC.