Hydraulics / Piping design

Liquid Cooling Pipe Sizing: Flow, Diameter and Pressure Loss

A pipe that carries enough liters per minute may still fail its pressure-drop budget. Calculate both before choosing nominal diameter.

Data center cooling infrastructure
Sections in this article

A four-rack GPU row is designed for 200 kW of heat in its technology cooling loop. The supplier quotes 360 L/min, and an installer chooses a 2½-inch header from the nominal size alone. That may be close to the flow requirement, but it does not verify the pipe's actual internal diameter, elbows, valve losses or available pump differential pressure.

Engineering order: determine how much heat the selected fluid must carry, calculate volume flow, choose a preliminary internal diameter against a velocity target, then evaluate friction and component losses against the pump curve. The Open Compute Project specifically highlights internal diameter—not nominal pipe size—as the quantity determining flow area, especially for non-metallic pipes with thicker walls.

Interactive liquid-cooling pipe estimator

Enter fluid properties for your specific coolant, not a universal “data-center water” value. The tool estimates theoretical heat-based flow, required diameter at a user-selected target velocity, and straight-pipe-only Darcy–Weisbach loss for the actual bore and stated combined pipe length. It estimates friction factor from Reynolds number and relative roughness. Results exclude fittings, strainers, CDUs, cold plates, manifolds and valves.

Hydraulic screeningFlow, internal diameter and friction
Required volumetric flow—
Diameter at target velocity—
Velocity at actual diameter—
Estimated straight-pipe pressure drop—

Preliminary straight-pipe model, not a pumping-system selection.

The equations and units that matter

Heat balanceQ = ṁ × cₚ × ΔT

Q is kW, mass flow ṁ is kg/s, and cₚ is kJ/(kg·K). Volume flow = mass flow divided by fluid density.

Pipe velocityv = 4 × V̇ / (π D²)

V̇ is m³/s and D is actual inside diameter in meters. Halving the bore produces four times the velocity at equal flow.

Darcy–WeisbachΔP = f × (L/D) × (ρv²/2)

Use Darcy friction factor f; it changes with Reynolds number and roughness. Pressure loss cannot be represented by v² alone when D and f change.

For turbulent flow, the estimator uses the explicit Swamee–Jain approximation for Darcy friction factor. For laminar flow it uses 64/Re. Reynolds numbers in the transitional region are flagged: an engineering calculation should not rely on a single smooth-pipe correlation there. Different fluids and temperatures change density, viscosity and heat capacity.

Worked example: four 50 kW racks, liquid-side duty

Assume all 200 kW are rejected into the chosen liquid loop, waterlike fluid at 995 kg/m³ and 4.18 kJ/(kg·K), and an 8°C temperature rise. Required flow is approximately 360.5 L/min. At a target 1.8 m/s, the calculated internal diameter is approximately 65.2 mm. A nominal trade size cannot be substituted for that value without checking its actual internal dimension.

Why this is a screening calculationA 65 mm internal bore gives approximately 1.81 m/s at 360.5 L/min.

A full pump selection additionally requires branch balancing, component pressure-drop curves, elevation effects in open systems, heat exchanger and strainer resistance, minimum chip-side flow, thermal expansion and the redundant-duty arrangement. Straight-pipe loss is only one part of total dynamic head.

If the same site uses glycol or a different secondary liquid, overwrite both cₚ and viscosity in the tool. Temperature-dependent properties should be evaluated at an appropriate operating point, not copied unchanged from room-temperature water. Check the fluid-vendor property data and material compatibility before release for construction.

Do not hide the pressure losses in fittings

Hydraulic elementWhat to obtainWhy it matters
Cold plates and rack manifoldsSupplier flow versus ΔP curves at temperatureSmall passages may dominate total resistance
Quick disconnects and hosesPressure loss at planned flow and permitted bend radiusLocalized restrictions, different connector sizes
Valves, elbows, filtersK values or equivalent lengths, dirty-filter allowanceOften omitted from first-pass estimates
CDU and pumpsPump curve, control range, duty/standby sequenceFlow cannot be established without available head
Pipe materialsInside diameter and compatible wetted materialsSame nominal size can have different flow area

A fixed velocity “maximum” is not universal. Velocity limits are a function of corrosion susceptibility, pipe material, noise, water chemistry, transient pressure and relevant standards. A calculator result should trigger engineering review, not automatically pronounce the design safe or unsafe.

Five checks before an installation drawing is approved

  1. Confirm the vendor's liquid-side portion of rack heat load; residual air heat is not automatically part of this circuit.
  2. Use the fluid's actual density, heat capacity, temperature and viscosity across the design range.
  3. Document the chosen pipe's inside diameter and pressure rating at maximum temperature.
  4. Build a hydraulic network model with fittings, valves, filters and equipment curves for each operating mode.
  5. Verify required flow at both design load and failed-pump conditions, then agree commissioning acceptance limits.

Save the estimator inputs beside the project drawing. Its value is making assumptions visible early, so that the engineer can replace estimates with the manufacturer's measured hydraulic data.

Sources and further reading

Primary frameworks and reference materials. Engineering decisions require site-specific validation; legal classifications require assessment of the actual use case.