Why Flow Rate Follows Directly from Heat Load and Temperature Rise
The flow rate a liquid cooling loop needs is not a free choice; it follows directly from two numbers: the heat load to be removed and the temperature rise the liquid is allowed to gain, through the sensible-heat equation that governs every liquid loop from a chilled-water plant to a direct-to-chip server cold plate.
The physics are straightforward. A liquid carries heat by warming up as it passes through the load. How much liquid you need depends on how much heat it must carry and how much it is allowed to warm. Carry more heat, or allow less temperature rise, and you need more flow. The relationship is Q = ṁ·cp·ΔT: the heat rate equals the mass flow times the specific heat times the temperature rise. Rearranged for flow, it says the required flow is the load divided by the product of specific heat and temperature rise. For water in Imperial units this collapses to the familiar 500 rule: gallons per minute equals BTU per hour divided by 500 times the temperature rise in Fahrenheit.
The calculator computes the required flow from the cooling load and the temperature rise, then normalizes it against the load as a flow intensity in gallons per minute per ton (or litres per second per kilowatt), and classifies that intensity from too low to too high. The normalized intensity is the screening metric: it reveals whether the chosen temperature rise is driving an unusually high or low circulation rate, independent of the loop size. This equation underlies chilled-water loops, process cooling, and the data-center liquid cooling that the Server Rack article pointed to when rack density climbs above the air-cooling ceiling of roughly 20 kW. Water properties anchor the calculation; glycol shifts the numbers.
Calculator Inputs: Cooling Load and Liquid Temperature Rise
The calculator uses three inputs and reports four outputs.
Unit System. Imperial (BTU/hr load input, GPM flow output, tons load reference, GPM/ton intensity) or Metric (kW load input, L/s flow output, L/s·kW intensity).
Cooling Load [BTU/hr or kW]. The heat rate the loop must remove. Taken from the equipment, rack, or coil thermal duty. Common references: 120,000 BTU/hr equals 10 tons equals 35.17 kW; 52 kW equals approximately 14.79 tons.
Liquid Temperature Rise [°F or °C]. The temperature the liquid is allowed to gain across the load from supply to return. The design choice that sets the flow. Typical chilled-water design: 10 to 16°F (5.5 to 8.9°C). Low delta-T runs (4 to 6°F) increase flow substantially.
The calculator outputs: flow rate in GPM or L/s; cooling load in tons or kW for reference; flow intensity in GPM/ton or L/s·kW; and a status badge classifying the intensity. The core relations:
Metric: flowRateLS = coolingLoad_kW / (4.186 × ΔT_C)
Imperial: flowRateGPM = coolingLoad_BTUh / (500 × ΔT_F)
or = (24 × tons) / ΔT_F
Flow intensity = flow / load [GPM/ton or L/s·kW]
Water-side model: the 4.186 (metric) and 500 (imperial) constants embed water's specific heat (4.186 kJ/kg·K, 1.0 BTU/lb·°F) and density (1000 kg/m³, 8.33 lb/gal). The calculator does not account for pressure drop, pump power, pump-curve compatibility, glycol or non-water fluid properties, velocity limits, two-phase cooling, equipment pressure limits, absolute supply temperature, or near-zero delta-T. It is a flow-basis screen, not a hydraulic design.
The Sensible-Heat Equation Behind the Flow Rate
The flow rate comes from the sensible-heat equation: the fundamental relation that a fluid's heat-carrying rate equals its mass flow times its specific heat times its temperature change.
Q = ṁ × cp × ΔT
where:
Q = heat rate removed [W or BTU/hr]
ṁ = mass flow rate [kg/s or lb/hr]
cp = specific heat of the liquid [kJ/kg·K or BTU/lb·°F]
ΔT = temperature rise across the load [°C or °F]
Solved for flow:
ṁ = Q / (cp × ΔT)
Convert mass flow to volume flow via density:
V̇ = ṁ / ρ [L/s or GPM]
For water in metric units, where cp equals 4.186 kJ/kg·K and density is approximately 1 kg/L, the volumetric form is:
flowRateLS = Q_kW / (4.186 × ΔT_C)
What each term does: the load Q is fixed by the equipment and drives flow proportionally. The specific heat cp is a fluid property; water's 4.186 kJ/kg·K is among the highest of common liquids, which is why water carries so much heat per unit volume and dominates cooling applications. The temperature rise ΔT is the design lever; increasing it reduces flow inversely (Section 6 addresses this trade-off in detail).
The equation describes sensible-heat transport. The liquid stays liquid and simply warms as it passes through the load. No phase change occurs. Two-phase cooling (boiling refrigerant, immersion with phase change) uses latent heat and a different equation. The sensible-heat equation is universal to single-phase liquid loops: it sizes a chilled-water plant, a process cooling loop, and a direct-to-chip data center circuit identically. Only the load, ΔT, and fluid differ.
Worked metric example:
52 kW, ΔT 6°C: flowRateLS = 52 / (4.186 × 6) = 52 / 25.12 = 2.07 L/s
Per ASHRAE Fundamentals: flow = Q / (cp·ΔT·ρ), the sensible-heat equation solved for flow. Water's high cp (4.186 kJ/kg·K) makes it the dominant coolant in single-phase loops.
The 500 Rule and the 24 Rule: The Imperial Shortcuts
In Imperial units the sensible-heat equation collapses into two shortcuts every HVAC engineer recognizes: the 500 rule and the 24 rule, both derived from water's properties.
The 500 rule:
GPM = BTU/hr / (500 × ΔT_F)
where 500 = 8.33 lb/gal × 60 min/hr × 1.0 BTU/lb·°F
The 500 constant assembles from water's density (8.33 lb/gal), the time conversion (60 min/hr), and water's specific heat (1.0 BTU/lb·°F). It converts a BTU/hr load and a °F temperature rise directly to GPM, with no intermediate unit conversions.
The 24 rule (when the load is stated in tons):
GPM = (24 × tons) / ΔT_F
where 24 = 12,000 BTU/hr per ton / 500
The 24 rule converts tons and degrees Fahrenheit to GPM directly. At the standard 10°F design temperature rise, it gives 2.4 GPM/ton (24 / 10 = 2.4). That "2.4 GPM/ton at 10°F" benchmark appears throughout HVAC practice as a memory anchor for conventional chilled-water design.
Both rules are water-only — they embed water's density and specific heat, so they do not apply to glycol or dielectric coolant loops without a correction factor. A 30% ethylene glycol mixture with a specific heat around 3.7 kJ/kg·K needs roughly 13% more flow than the 500 rule gives; the same 500 constant understates flow for a glycol loop.
Worked Imperial example:
120,000 BTU/hr, 10°F: GPM = 120,000 / (500 × 10) = 120,000 / 5,000 = 24.0 GPM
or: (24 × 10 tons) / 10°F = 240 / 10 = 24.0 GPM (same result)
Per ASHRAE HVAC Systems and Equipment: the 500 rule (GPM = BTU/hr / (500·ΔT)) and 24 rule (GPM = 24·tons / ΔT) are the Imperial water-side shortcuts, embedding 8.33 lb/gal × 60 × 1.0 BTU/lb·°F. Water-only; glycol changes the constants.
Temperature Rise Drives Flow: The Inverse Relationship
The temperature rise is the design lever, and because flow is inversely proportional to it, a small change in the chosen ΔT swings the required flow substantially.
flow ∝ 1/ΔT
Halve the ΔT → double the flow
Double the ΔT → halve the flow
The trade-off between low and high ΔT:
Low ΔT (e.g. 4°F): high flow, more pumping energy, tighter temperature control, more margin
High ΔT (e.g. 16°F): low flow, less pumping energy, larger temperature swing, less margin
The load is fixed by the equipment being cooled; the designer cannot change it. The ΔT is the variable the engineer chooses, and that choice fully determines the required flow. Choosing a lower ΔT means circulating more fluid to remove the same heat with a smaller temperature change. Choosing a higher ΔT means circulating less but warming the fluid more across the load.
Normalized intensity depends on ΔT alone for a given fluid. For water in Imperial units, GPM/ton = 24 / ΔT_F. At 10°F, intensity is 2.4 GPM/ton regardless of whether the loop serves 5 tons or 500 tons. This means the normalized metric screens the ΔT choice directly, stripping out loop size and leaving only the design temperature.
The pumping energy angle is significant for large chilled-water systems. Low ΔT equals high flow equals high pump power. Per ASHRAE 90.1, higher ΔT designs reduce circulating flow and pump energy, and 90.1 consequently favors higher chilled-water ΔT in energy code compliance. A 16°F ΔT design needs half the flow of an 8°F ΔT design on the same load, cutting pump power by more than half when the system sees the cubic relationship between flow and power at a fixed system curve.
Worked contrast:
10 tons at 10°F: 24 GPM (2.4 GPM/ton, RECOMMENDED)
10 tons at 4°F: 60 GPM (6.0 GPM/ton, TOO HIGH, excess pumping)
10 tons at 20°F: 12 GPM (1.2 GPM/ton, TOO LOW, thin margin)
Per ASHRAE 90.1 and Fundamentals: flow is inversely proportional to ΔT. Low ΔT means high flow (pumping energy, more margin); high ΔT means low flow (efficiency, less margin). ΔT is the design lever; higher-ΔT designs save pump energy and are favored by energy codes.
Normalized Flow Intensity: Why GPM per Ton Screens the Design
The calculator's key output is not the raw flow but the normalized flow intensity: flow per unit of cooling. That single number screens whether the temperature-rise choice is reasonable regardless of loop size.
Flow intensity = flow / load
Imperial: GPM/ton
Metric: L/s·kW
Normalizing removes loop size from the comparison. A 5-ton loop and a 500-ton loop designed at the same ΔT have the same GPM/ton: raw GPM differs by a factor of 100, but GPM/ton is identical because both serve proportionally sized loads. Raw GPM conceals whether the design is reasonable; GPM/ton exposes the temperature-rise choice directly.
For water, the intensity depends on ΔT alone (from the 24 rule):
GPM/ton = 24 / ΔT_F
10°F → 2.4 GPM/ton
8°F → 3.0 GPM/ton
12°F → 2.0 GPM/ton
4°F → 6.0 GPM/ton
What the intensity reveals: a high intensity (above 3 to 4 GPM/ton) signals that the ΔT is too low, forcing excess flow and pumping energy. A low intensity (below 1.5 to 2.0 GPM/ton) signals that the ΔT is too high, leaving thin thermal margin if the load peaks above design. The recommended band of 2 to 3 GPM/ton corresponds to the conventional 8 to 12°F chilled-water ΔT range and represents the balance between margin and efficiency that ASHRAE practice identifies as sound design.
The trap with raw flow: a large system can have high total GPM yet a perfectly acceptable GPM/ton. A 200-ton chilled-water plant at 2.4 GPM/ton runs 480 GPM, which sounds large; the intensity confirms the design is conventional. Do not judge by total flow; judge by intensity relative to load.
Per ASHRAE practice: normalized flow intensity (GPM/ton or L/s·kW) screens the design by removing loop size, isolating the ΔT choice. For water, GPM/ton = 24 / ΔT_F. Judge intensity, not raw flow.
The Recommended Band and Its Interpretation Tiers
The calculator maps flow intensity to five tiers, with the recommended band at 2 to 3 gallons per minute per ton, so a computed intensity becomes a judgment on the design.
| Flow Intensity (GPM/ton) | L/s·kW | Status |
|---|---|---|
| < 1.5 | < 0.026 | TOO LOW |
| 1.5 to < 2.0 | 0.026 to 0.035 | LOW / MARGINAL |
| 2.0 to 3.0 | 0.035 to 0.053 | RECOMMENDED |
| > 3.0 to 4.0 | 0.053 to 0.070 | HIGH |
| > 4.0 | > 0.070 | TOO HIGH |
Tier by tier: below 1.5 GPM/ton corresponds to a ΔT above roughly 16°F, leaving thin thermal margin and risk if the load peaks above design. The 1.5 to 2.0 band (ΔT 12 to 16°F) is approaching thin and warrants review. The recommended 2.0 to 3.0 GPM/ton (ΔT 8 to 12°F) is the practical chilled-water design range per ASHRAE HVAC Systems. The 3.0 to 4.0 band (ΔT 6 to 8°F) is high but acceptable when a low ΔT is intentional: for tight temperature control, direct-to-chip applications, or where the supply-return temperature spread is constrained by the heat exchanger. Above 4.0 GPM/ton (ΔT below 6°F), the flow and pumping energy are excessive; raising the ΔT is the standard corrective action.
The band in ΔT terms:
2.0 GPM/ton ≈ 12°F ΔT (or 6.7°C)
3.0 GPM/ton ≈ 8°F ΔT (or 4.4°C)
The recommended 2-3 GPM/ton band = the 8-12°F chilled-water design range
The thresholds are inclusive at the boundaries: 2.0 and 3.0 GPM/ton classify as RECOMMENDED; 4.0 GPM/ton is HIGH. Data-center direct-to-chip loops and low-ΔT process applications may intentionally run in the HIGH tier. The band screens conventional chilled-water design; context determines whether a tier-4 design is acceptable.
Per ASHRAE HVAC Systems: recommended flow intensity is 2 to 3 GPM/ton (0.035 to 0.053 L/s·kW), corresponding to an 8 to 12°F chilled-water ΔT. Below 1.5 is thin margin; above 4.0 is excess pumping. The band screens conventional designs; context governs.
Glycol Changes the Numbers: Specific Heat and Density
The calculator uses water properties, but glycol antifreeze mixtures have a lower specific heat and higher density, so a glycol loop needs more flow than water for the same load and ΔT, and the water-side result requires correction.
The property shifts at typical concentrations:
Water (pure): cp 4.186 kJ/kg·K, ρ 1000 kg/m³
Ethylene glycol 30%: cp ~3.7 kJ/kg·K (lower), ρ ~1030 kg/m³ (higher)
Propylene glycol 30%: cp ~3.8 kJ/kg·K (lower), ρ ~1025 kg/m³ (higher)
Because flow is inversely proportional to cp, a lower specific heat requires proportionally more flow to carry the same heat. Ethylene glycol at 30% concentration has a cp roughly 12% below water; the glycol loop needs approximately 12% more flow than the water-side result for the same load and ΔT. Higher concentrations lower cp further and compound the flow penalty.
The correction:
flow_glycol = flow_water × (cp_water / cp_glycol) × (ρ correction)
Apply manufacturer glycol property data at the design concentration and temperature.
When glycol is used: outdoor loops and economizer coils exposed to freezing, chilled-water systems in cold climates, some data-center secondary loops serving coolant distribution units. The penalty stacks: lower cp demands more flow; higher viscosity raises pressure drop and pump power; lower heat-transfer coefficient may require a larger heat exchanger surface. Do not use the raw water result for a glycol loop without the correction factors.
Per ASHRAE Fundamentals and glycol manufacturer data (Dow Chemical DOWFROST/DOWTHERM tables): glycol has lower specific heat (needs more flow) and higher density and viscosity (more pumping) than water. A glycol loop needs approximately 10 to 20% more flow than the water-side result; apply correction factors at the design concentration and operating temperature.
Liquid Cooling in Data Centers: Where the Density Ladder Ends
Liquid cooling is where the data-center density ladder ends, because above roughly 20 kilowatts per rack air can no longer carry the heat, and the same sensible-heat flow equation then sizes the coolant loop to the chip.
The density ladder (per the Server Rack article):
< 10 kW/rack: room CRAC (air)
10-20 kW/rack: in-row / rear-door heat exchangers (air, local)
> 20 kW/rack: liquid (direct-to-chip or immersion)
Air cooling ends near 20 kW per rack because air has low density and a low specific heat (cp approximately 1.0 kJ/kg·K, compared with water's 4.186). Above that density threshold, the airflow volume needed to remove rack heat would saturate supply capacity or require impractical inlet velocities. Water carries roughly four times the heat per unit mass and far more per unit volume, making it the practical coolant for high-density racks.
Liquid cooling architectures for data centers:
Direct-to-chip (cold plate): coolant loop to a plate on the CPU/GPU; most common for 30-80 kW racks
Rear-door heat exchanger: coil on the rack back, air-to-liquid; serves the 15-25 kW middle range
Immersion: servers submerged in dielectric fluid; emerging for highest density
The sensible-heat equation still governs. A direct-to-chip loop removing 40 kW at a 10°C rise:
flowRateLS = 40 / (4.186 × 10) = 40 / 41.86 = 0.956 L/s (15.2 GPM)
The same Q = ṁ·cp·ΔT, applied to the chip level. A coolant distribution unit (CDU) typically separates the facility water loop from the secondary technology-cooling loop; each loop sizes independently by the same equation at its own ΔT.
Per ASHRAE TC 9.9 liquid cooling guidelines and Open Compute Project Open Rack v3: liquid cooling begins above approximately 20 kW per rack, where air's low specific heat fails. The sensible-heat equation sizes the coolant loop at the chip. GPU and AI racks drawing 30 to 80 kW (NVIDIA H100/H200 designs) are firmly in liquid territory — accelerating the mainstream adoption of direct-to-chip cooling.
Worked Example: 10 Tons at a 10-Degree Rise to 24 GPM
This example matches the calculator's Imperial worked example: a water-based loop serving a 10-ton cooling load at a 10°F design temperature rise.
Step 1. Confirm the load:
10 tons × 12,000 BTU/hr per ton = 120,000 BTU/hr (35.17 kW)
Step 2. Apply the 500-rule flow:
GPM = 120,000 / (500 × 10) = 120,000 / 5,000 = 24.0 GPM (1.512 L/s)
Step 3. Verify with the 24 rule:
GPM = (24 × 10 tons) / 10°F = 240 / 10 = 24.0 GPM (same result)
Step 4. Metric cross-check:
coolingLoad_kW = 120,000 × 0.000293071 = 35.17 kW
ΔT_C = 10 × 0.55556 = 5.556°C
flowRateLS = 35.17 / (4.186 × 5.556) = 35.17 / 23.26 = 1.512 L/s
GPM = 1.512 × 15.8503 = 23.97 ≈ 24.0 GPM (consistent)
Step 5. Normalized flow intensity:
GPM/ton = 24.0 / 10.0 = 2.40 GPM/ton
Step 6. Classification:
2.40 GPM/ton falls in the 2.0-3.0 range → RECOMMENDED
Step 7. Interpretation: 24 GPM of water removes 10 tons at a 10°F rise. A 2.40 GPM/ton intensity is a conventional chilled-water result. The 10°F ΔT sits in the recommended 8 to 12°F range.
Step 8. ΔT sensitivity at the same 10-ton load:
8°F: GPM = (24 × 10) / 8 = 30.0 GPM (3.0 GPM/ton, RECOMMENDED, higher pumping)
12°F: GPM = (24 × 10) / 12 = 20.0 GPM (2.0 GPM/ton, RECOMMENDED, lower pumping)
5°F: GPM = (24 × 10) / 5 = 48.0 GPM (4.8 GPM/ton, TOO HIGH)
Step 9. Glycol note: if the loop uses 30% ethylene glycol, apply the cp correction. The adjusted flow is approximately 10 to 15% higher (about 27 GPM). Do not use the 24 GPM water result directly for a glycol loop.
Step 10. Result and next step: 24.0 GPM, 2.40 GPM/ton, RECOMMENDED. Conventional 10°F chilled-water design. Next: verify pressure drop, velocity, and pump power from this flow basis. Cross-reference HVAC Delta T for the temperature choice, Chiller Capacity for the plant, Server Rack for density context.
Metric and Too-High Worked Examples
Two further scenarios match the calculator's metric and high-flow worked examples.
Metric example (52 kW process or data-center loop at 6°C rise):
Step 1. Load + ΔT: 52 kW, 6°C temperature rise
Step 2. Flow rate:
flowRateLS = 52 / (4.186 × 6) = 52 / 25.12 = 2.07 L/s (32.8 GPM)
Step 3. Normalized intensity:
flowIntensityMetric = 2.07 / 52 = 0.0398 ≈ 0.040 L/s·kW
Step 4. Classification:
0.040 L/s·kW falls in 0.035-0.053 → RECOMMENDED
Step 5. Interpretation: 2.07 L/s removes 52 kW at a 6°C rise.
0.040 L/s·kW is a conventional intensity. 6°C ≈ 10.8°F, within the recommended range.
Too-high example (8 tons at a 4°F rise: excess flow from very low ΔT):
Step 6. Load + low ΔT: 8 tons (96,000 BTU/hr), 4°F temperature rise
Step 7. Flow rate:
GPM = (24 × 8) / 4 = 192 / 4 = 48.0 GPM (3.03 L/s)
Step 8. Normalized intensity:
GPM/ton = 48.0 / 8.0 = 6.0 GPM/ton
Step 9. Classification:
6.0 GPM/ton > 4.0 → TOO HIGH
Step 10. Interpretation and corrective action:
A 4°F ΔT forces 6.0 GPM/ton, far above the recommended band.
48 GPM for 8 tons is excessive: oversized piping, oversized pump, high pumping energy.
Corrective action: raise the ΔT to 10°F → 19.2 GPM, 2.4 GPM/ton, RECOMMENDED.
Per ASHRAE: the metric case (52 kW, 6°C) gives 2.07 L/s, 0.040 L/s·kW, RECOMMENDED. The too-high case (8 tons, 4°F) gives 48 GPM, 6.0 GPM/ton, TOO HIGH. Low ΔT is the standard cause of excess flow intensity; raising the temperature rise corrects it.
Application Boundaries: Pressure Drop, Pump Power, Fluid Limits, Two-Phase
The calculator computes water-side sensible-heat flow from load and ΔT and normalizes the result as a flow intensity. Several analyses fall outside that scope and require separate calculation.
Pressure drop. The calculator gives flow, not pressure drop through the loop (pipe, coil, valves, cold plate, heat exchanger). Hydraulic sizing requires a separate pressure-drop calculation across each component and circuit.
Pump power. Flow is one input to pump power; head (pressure drop) is the other. Pump power = flow × head / efficiency, a separate calculation. The flow basis from this calculator feeds that analysis.
Pump-curve compatibility. The required flow must land on a real pump's operating curve at the system head. Pump selection follows pump-power and pressure-drop calculation.
Glycol and non-water fluids. The model uses water properties. Glycol, dielectric immersion fluids, and other coolants need cp, ρ, and viscosity corrections (Section 9 covers glycol; immersion fluids differ more substantially).
Velocity and erosion limits. High flow rates raise pipe velocity, risking erosion and noise. Typical velocity limits are 8 to 10 ft/s (2.4 to 3.0 m/s) in copper pipe. The calculator does not verify velocity; check against the design pipe size.
Two-phase and refrigerant cooling. The sensible-heat equation applies to single-phase liquid only. Two-phase cooling (refrigerant evaporation, boiling immersion, phase-change materials) uses latent heat and requires a different equation.
Equipment pressure limits. Cold plates, heat exchangers, and CDUs have maximum operating pressure and flow limits. These are not checked by the flow calculator.
Absolute supply temperature. The model uses only the supply-to-return ΔT, not the absolute supply temperature. The supply temperature (chilled-water setpoint, ASHRAE TC 9.9 facility water class) affects coil capacity and dew-point risk, requiring separate analysis.
Zero or near-zero ΔT. Division by a very small ΔT produces an absurd flow result. The ΔT must be a real design value, not a limit case.
Fouling and degradation. The model assumes clean design conditions. Fouling raises effective flow requirements over time and requires design margin.
Per ASHRAE Fundamentals: water-side sensible-heat flow and intensity screening are the calculator scope. Pressure drop, pump power, pump-curve selection, glycol correction, velocity limits, two-phase cooling, equipment limits, and absolute supply temperature require separate analysis. A qualified engineer completes the hydraulic and pump design from this flow basis.
Liquid Cooling Flow Rate Calculator
Liquid cooling flow rate from the sensible-heat equation: divides the cooling load by the product of the liquid's specific heat and the design temperature rise, Q = ṁ·cp·ΔT solved for flow, giving gallons per minute or litres per second. For water this is the 500 rule (GPM = BTU/hr / (500·ΔT)) or the 24 rule (GPM = 24·tons / ΔT). It normalizes the result as a flow intensity in GPM per ton (or L/s·kW), screening it against the recommended 2 to 3 band. Water-side properties; apply glycol correction for antifreeze loops. A flow-basis screen, not a pressure-drop or pump-power calculation.
FAQ
How do you calculate liquid cooling flow rate?
Per ASHRAE Fundamentals: from the sensible-heat equation, flow = load / (specific heat × temperature rise × density). For water in Imperial units this is the 500 rule: GPM = BTU/hr / (500 × ΔT°F), where 500 embeds water's 8.33 lb/gal density and 1.0 BTU/lb·°F specific heat. More load or less ΔT means more flow, proportionally. In metric units: L/s = kW / (4.186 × ΔT°C).
What is the 500 rule in HVAC?
Per ASHRAE HVAC Systems and Equipment: GPM = BTU/hr / (500 × ΔT°F), where 500 = 8.33 lb/gal × 60 min/hr × 1.0 BTU/lb·°F, assembling water's density, time conversion, and specific heat into a single water-side constant. It gives chilled-water flow directly from load and temperature rise. The equivalent tons form is GPM = 24 × tons / ΔT, where 24 = 12,000 BTU/hr per ton / 500.
What is a good GPM per ton for chilled water?
Per ASHRAE practice: 2 to 3 GPM/ton is the recommended band, corresponding to a design temperature rise of 8 to 12°F (4.4 to 6.7°C). Below 1.5 GPM/ton signals a ΔT above 16°F, leaving thin thermal margin. Above 4.0 GPM/ton signals a ΔT below 6°F, producing excess flow and pumping energy. At the standard 10°F ΔT, water circulates at 2.4 GPM/ton.
Why does lower temperature rise need more flow?
Per the sensible-heat equation: flow is inversely proportional to ΔT. The liquid must carry the same heat load with a smaller temperature change, so more of it must circulate per unit time. Halving the ΔT doubles the required flow and can more than double pumping energy, because pump power also rises with the increased pressure drop in the larger-flow circuit. This inverse relationship is why ASHRAE 90.1 favors higher-ΔT designs for energy efficiency.
Does glycol change the flow rate?
Per ASHRAE Fundamentals and glycol manufacturer data: yes. Glycol has a lower specific heat than water (ethylene glycol 30% is approximately 3.7 kJ/kg·K, versus water's 4.186), so a glycol loop needs more flow to carry the same heat at the same ΔT. The required increase is roughly 10 to 20% at typical concentrations, plus a further penalty from higher viscosity, which raises pressure drop. Apply correction factors at the design concentration; do not use the water-side result directly.
When does a data center need liquid cooling?
Per ASHRAE TC 9.9: above approximately 20 kW per rack, where air's low specific heat (1.0 kJ/kg·K versus water's 4.186) cannot remove rack heat without severe inlet temperature violations. Direct-to-chip (cold plate), rear-door heat exchangers, and immersion are the three liquid architectures, each sized by the same Q = ṁ·cp·ΔT equation at the rack or chip level. GPU and AI racks drawing 30 to 80 kW are firmly in liquid territory.
Does this calculate pump power or pressure drop?
Per ASHRAE: no. The calculator gives the required flow and its normalized intensity as a flow-basis screen. Pressure drop (hydraulic loss through the pipe, coil, heat exchanger, and fittings) and pump power (flow × head / efficiency) are separate calculations needed to complete the hydraulic design and pump selection. The flow from this calculator is the starting point for those analyses.
Related Calculators
- Server Rack Heat Load Calculator: The per-rack density ladder whose top rung (above 20 kW) this liquid flow serves (article).
- Data Center CRAC Redundancy Calculator: Cooling redundancy for the air side below the liquid threshold (article).
- HVAC Delta T Calculator: The supply-return temperature difference that drives this flow rate.
- Chiller Capacity Calculator: The chilled-water plant capacity behind the loop.
- Glycol Concentration Calculator: Antifreeze concentration that shifts the specific heat and flow correction.
- Hydronic Balancing Calculator: Balancing flow across the loop's circuits.
- Cooling Load Calculator: The space cooling load feeding the loop duty.
- Fan Power Calculator: Air-side power for comparison with liquid pumping.
References
- ASHRAE. ASHRAE Handbook: HVAC Systems and Equipment, 2020 ed. ASHRAE, Atlanta, GA. (chilled-water system design, the 500 rule, the 24 rule, flow intensity bands, GPM/ton recommended band)
- ASHRAE. ASHRAE Handbook: Fundamentals, 2021 ed. Chapter 1: Thermodynamics; Chapter 3: Fluid Flow. ASHRAE, Atlanta, GA. (sensible-heat equation Q = ṁ·cp·ΔT, flow = Q/(cp·ΔT·ρ), water cp 4.186 kJ/kg·K, single-phase liquid transport)
- ASHRAE. ASHRAE Handbook: HVAC Applications, 2019 ed. Chapter 48: Data Centers. ASHRAE, Atlanta, GA. (liquid-side loop design, chilled-water applications, flow rate recommendations for data-center loops)
- ASHRAE Standard 90.1-2022, Energy Standard for Buildings Except Low-Rise Residential Buildings. ASHRAE, Atlanta, GA. (chilled-water ΔT requirements, pump energy, higher-ΔT efficiency incentive, Sections 6.5 and 6.8)
- ASHRAE Technical Committee TC 9.9. Thermal Guidelines for Data Processing Environments, 4th ed. ASHRAE, Atlanta, GA, 2015. (liquid cooling above 20 kW/rack, facility water classes W17-W45, direct-to-chip and immersion architectures, flow sizing at chip level)
- Open Compute Project. Open Rack v3 Hardware Specification and Liquid Cooling Guidelines. OCP Foundation, 2022. (coolant distribution unit flow sizing, secondary loop design, direct liquid cooling for hyperscale racks)
- Dow Chemical Company. DOWFROST and DOWTHERM SR-1 Engineering Data: Glycol Fluid Properties. Dow Chemical, Midland, MI. (ethylene and propylene glycol cp, ρ, viscosity vs. concentration and temperature; correction factors for HVAC loop sizing)
- NVIDIA Corporation. H100/H200 Tensor Core GPU Thermal Reference Design and Liquid Cooling Specifications. NVIDIA, Santa Clara, CA, 2023-2024. (GPU rack power 30-80 kW, direct-to-chip flow requirements, CDU secondary loop sizing)
- ASHRAE. ASHRAE Handbook: Refrigeration, 2022 ed. Chapter 2: Fluid Flow. ASHRAE, Atlanta, GA. (single-phase sensible-heat transport vs. two-phase latent heat; phase-change equation distinctions)
- International District Energy Association (IDEA). District Cooling Best Practices Guide. IDEA, Westborough, MA, 2021. (flow intensity ranges in district cooling, GPM/ton benchmarks, ΔT optimization for large chilled-water plants)
- ASHRAE. ASHRAE Handbook: HVAC Applications, 2019 ed. Chapter 3: Commercial and Public Buildings. ASHRAE, Atlanta, GA. (process cooling loop flow sizing, chilled-water loop design for commercial applications)
- Carrier Corporation. Handbook of Air Conditioning System Design. McGraw-Hill, New York, NY. (500 rule derivation, 24 rule application, chilled-water loop sizing methodology, water-side flow constants)