Helium Leak Testing for Liquid Cooling Components: Methods, Leak Rates and What to Specify

A cold plate passes its pressure decay test at the factory. Three months after it goes into an AI server rack, a technician finds a ring of dried coolant around one of the port welds. When that happens, the problem usually isn't that nobody tested the part. It's that the test wasn't sensitive enough, or the acceptance limit didn't match the risk. Helium leak testing exists to close that gap.

Inside a liquid-cooled rack, a small leak can corrode a board, starve a loop of flow, or take a whole row offline, and the resulting outage can cost hundreds of thousands of dollars an hour. That's why more server OEMs and system integrators no longer write "leak test required" on a drawing. They specify helium leak testing in detail: the leak rate, test pressure and test method.

This guide looks at helium leak testing from the buyer's and engineer's side: how it works, the main methods, the standards behind them, how to set a leak rate limit, and what to put in your RFQ so suppliers quote on the same basis and ship parts you can actually accept.

What Is Helium Leak Testing?

Helium leak testing, also called a helium leak check, is a nondestructive method that uses helium as a tracer gas to find and measure leaks. The part is pressurized with helium or evacuated and exposed to it, and a mass spectrometer leak detector measures how much helium passes through, reported as a leak rate in mbar·L/s.

For cold plates, manifolds and quick disconnects, it's the most sensitive method in common production use for proving a part is sealed before it ships.

How Helium Leak Testing Works

Helium makes a good tracer gas for three reasons. Its atoms are small enough to pass through extremely fine leak paths. It's inert, nonflammable and nontoxic, so it won't react with copper, aluminum, stainless steel or seal materials. And it makes up only about 5 ppm of the air, so nearly all the helium a detector sees comes from the leak, not the room.

One caveat matters for liquid cooling assemblies: elastomers let helium permeate directly through the material. On a manifold or QD assembly sealed with O-rings, that permeation can raise the background signal over time and look like a leak, so the test time and background handling need to be agreed with your supplier.

The instrument at the center of the test is the mass spectrometer leak detector (MSLD). It counts helium ions at mass 4 and converts that count into a leak rate.

The unit, mbar·L/s, is easier to picture than it sounds: a leak rate of 1 mbar·L/s means the pressure in a 1-liter vessel changes by 1 mbar every second. Rates of practical interest run from around 10⁻¹ (a visible drip) down to 10⁻⁹ and below. For related terms, see our liquid cooling glossary.

Specs from different customers and regions use different units. Convert to one unit before you compare quotes or test reports.

Why Pressure Decay Alone Isn't Enough for Liquid Cooling Parts

Pressure decay testing fills a part with compressed air, isolates it and watches for a pressure drop. It's cheap, fast and still one of the most common leak tests for cold plates and manifolds. But it has two built-in limits: its sensitivity drops as internal volume grows, and temperature swings can skew the reading. On a large GPU cold plate or a multi-branch manifold, pressure decay struggles to resolve the very small leak rates that a helium test can measure.

Yet many liquid cooling leaks are that small, and they tend to form in the same places:

• Cold plates: brazed or friction stir welded seams, cover sealing faces, inlet and outlet fittings, and small machining defects. For how each joining process affects sealing, see our cold plate manufacturing process guide.
• Manifolds: multiple branch connections, O-rings and valve blocks, all under thermal expansion and contraction in service.
• Quick disconnect assemblies: thread seals, O-rings and the QD itself are all potential leak paths. A cold plate that passes on its own won't necessarily pass once the QDs go on. Our quick disconnect guide covers how to handle that.

So a common approach uses two stages: screen out gross leaks with pressure decay or a bubble test, then use helium to catch the fine ones.

Helium Leak Test Methods

The four main helium leak test methods, sometimes called helium leak detection methods, are vacuum chamber testing, sniffer testing, spray testing and helium bombing. They fall into two groups based on which way the tracer gas moves: helium leaks into the part (vacuum, or outside-in, methods) or out of it (pressure, or inside-out, methods). ISO 20485 calls these Group A and Group B techniques. These are the ones you'll see most with liquid cooling hardware.

Vacuum Chamber (Hard Vacuum) Testing

The part goes into a vacuum chamber and is filled with helium, and then the chamber is pumped down. The leak detector measures all the helium that escapes from the part into the chamber. Despite the name, this is an inside-out test: the helium flows out of the part. It's an integral test: you get the total leak rate for the whole part, but not the location of the leak.

Its strengths are high sensitivity, a quantified result and easy automation. Hard vacuum testing can detect leaks from 10⁻⁴ to 10⁻⁸ atm·cc/s and beyond (1 atm·cc/s is roughly 1 mbar·L/s), which makes it a good fit for production acceptance of cold plates and manifolds. The trade-offs are the cost of the chamber and tooling, and a longer cycle time than pressure decay.

Sniffer (Detector Probe) Testing

In helium sniffing, the part is pressurized with helium, and an operator or robot moves a sniffer probe along the welds and fittings. The probe draws any escaping helium into the detector. ASTM E499 covers this method and describes two ways to run it:

• Direct probing: used mainly to locate leaks. It gives only an approximate leak rate, and results depend on the background helium level and how much of the leak the probe captures.
• Accumulation: the part or a section of it is enclosed so escaping helium builds up. Comparing the signal against a calibrated leak gives a measured leak rate.

It's less sensitive than chamber testing. Industrial sniffer leak detectors reach down to about 2×10⁻⁷ mbar·L/s, while helium vacuum systems detect leaks in the 10⁻¹² mbar·L/s range. But sniffing doesn't need a vacuum chamber. That makes it useful for parts too large for a chamber, and for pinpointing the leak on a part that failed a chamber test.

Spray (Outside-In) Testing

The inside of the part is evacuated and connected to the detector, and helium is sprayed on the outside, one spot at a time. Wherever there's a leak, helium is drawn in and detected. It's the vacuum-side way to locate leaks, and it's sensitive, but it isn't well suited to integral acceptance testing at volume.

Helium Bombing

The part soaks in a pressurized helium tank so helium can enter any leak paths. It's then moved to a vacuum chamber for measurement. Bombing is meant for fully sealed devices that can't be filled directly, such as relays and electronic packages. Cold plates and manifolds have ports and can be filled directly, so bombing is rarely used for them.

Method Comparison

Helium Leak Testing Standards

Citing a standard in your RFQ or drawing is far clearer than writing "helium test." The ones you'll see most often:

• ASTM E493: mass spectrometer testing in inside-out mode, with two methods, bombing and prefilling. The OCP quick disconnect specification requires every helium leak test to follow ASTM E493.
• ASTM E499: detector probe (sniffer) mode, covering direct probing and accumulation.
• ASTM E498: tracer probe mode, where the part is evacuated and helium is applied to the outside to locate leaks.
• ASTM E1603: hood mode, where the evacuated part is enclosed in a helium-filled hood to measure its total leak rate.
• ISO 20485: the tracer gas method. It replaced Europe's EN 13185.
• ISO 20484: leak testing vocabulary.
• ISO 20486: calibration of reference leaks, which keeps detector readings traceable.
• EN 1779: criteria for selecting a leak testing method and technique.

Note that ASTM E493 is written for parts sealed before testing. For open-port parts such as cold plates and manifolds, ISO 20485 covers the full range of tracer gas techniques, including vacuum chamber, sniffing and accumulation. Whichever you cite, a standard tells you how to test. It doesn't tell you what result passes. The buyer still has to set the leak rate limit for the application.

How Tight Is Tight Enough? Setting a Helium Leak Rate

There is no single correct helium leak rate for liquid cooling parts. The limit should be set from the consequences of a leak, the operating pressure and the coolant, and it only means something when the test pressure, helium concentration and method are stated with it. This is the question buyers ask most, and the line they most often get wrong.

Gas Leak Rate vs. Coolant Leakage

Helium testing measures gas, but what you actually care about is coolant. There's no universal conversion between the two, because whether a liquid gets through depends on the size of the leak path, the surface material and the liquid's surface tension. Depending on the materials, "watertight" can mean anything from 10⁻² to 10⁻⁵ mbar·L/s, which is why "watertight" doesn't work as an acceptance criterion.

INFICON ran a water drip test on artificial leaks that helps build intuition. Here is what it observed over 30 minutes at a 100 mbar pressure difference:

Keep the conditions in mind: this was a low-pressure IP67 water ingress test at 100 mbar. Liquid cooling loops run at several bar (the OCP quick disconnect specification, for example, sets a maximum operating pressure of 6.89 bar(g)), so the same leak path would pass more liquid in service. Use the table for orders of magnitude, not as a cold plate acceptance limit.

A Published Industry Reference

The closest published number for liquid cooling hardware comes from the OCP MQD/MQDB Design Specification (revision 1.0.0, approved October 2026). QDs are helium leak tested before and after temperature cycling and connect-disconnect cycling tests, and the leak rate must be 1×10⁻⁵ mbar·L/s or lower.

Its scope is narrow, though. It applies to quick disconnects, and it's a qualification test run on small sample sets, not on every part. It isn't a production acceptance limit for cold plates, and it isn't a 100% inspection requirement. Use it as a reference point when you set your own limit, not as a value to copy onto a cold plate drawing.

Guidelines for Setting the Limit

1. Start with consequences. What happens if this part leaks? Is it near live electronics? Is the coolant water-based or dielectric? Can it be serviced in place? The worse the outcome, the tighter the limit.
2. Don't write "zero leak." Every method has a detection limit, so zero leakage can't be verified. All it does is let each supplier interpret the spec, and price it, their own way.
3. State the test conditions. A leak rate means little without the test pressure, helium concentration and method. The same part reads differently with 100% helium vs. 10% helium, or at 3 bar vs. 10 bar.
4. Tighter costs more. A stricter limit means more expensive equipment, longer cycle times and higher scrap. Match it to the real risk.

Helium vs. Pressure Decay, Bubble and Hydrogen Tracer Testing

Forming gas, typically 95% nitrogen and 5% hydrogen, is increasingly used as a lower-cost alternative to helium. Before you switch, confirm it can reach the leak rate you need and that it's compatible with the part's materials.

A Typical Helium Leak Test Procedure

Equipment and fixtures vary from shop to shop, but helium testing of liquid cooling parts generally follows this sequence:

1. Clean and dry the part. Leftover water, cutting fluid or cleaning agents can plug fine leak paths and hide a leak, so the part must be dry before testing.
2. Run a gross leak check. Screen out large leaks with pressure decay or a bubble test first. A gross leak can flood the helium detector, raise its background for the next parts and drag out pump-down time.
3. Evacuate and fill with helium. For chamber testing, evacuate the part, fill it with helium at the specified concentration and pressure, then pump down the chamber. For sniffing, pressurize the part with helium.
4. Take the reading. Once the detector stabilizes, read the leak rate and subtract the background.
5. Verify with a calibrated leak. At the specified interval, check the detector against a reference leak of known rate so results stay traceable.
6. Judge and record. Compare the reading with the acceptance limit, and record the serial number, leak rate, test pressure and time. Use sniffing or spray testing to find the leak on any rejected part.

How to Specify Helium Leak Testing in Your RFQ or Drawing

A helium leak test specification should state the test method, the acceptance limit with units, the test pressure and helium concentration, the test stage, the sampling plan, where the test falls in the test sequence, calibration, post-test handling and reporting. Our custom cold plate RFQ guide lists everything a complete RFQ should include. For helium leak testing specifically, cover at least the following:

• Test method: vacuum chamber, sniffer or accumulation, plus the standard you're citing (such as ISO 20485 or ASTM E499).
• Acceptance limit and units: for example, "≤ X mbar·L/s (helium)." Write out the units, and state whether the value is a helium rate or an air-equivalent rate.
• Test pressure and helium concentration: the internal helium pressure, the concentration and the condition of the part during the test.
• Test stage: the bare cold plate, the part after fittings are joined, or the full assembly with QDs installed.
• Sampling: 100% inspection or lot sampling in production, and any separate requirements for first articles and qualification.
• Test sequence: where the helium test falls relative to proof, burst and thermal cycling tests. A proof test can open a micro-leak in a joint that passed earlier, so a common practice is to leak test again after proof testing.
• Calibration: how often reference leaks are calibrated and what traceability you require.
• Post-test handling: drying and cleanliness requirements, and whether parts ship with a nitrogen purge or port caps.
• Reporting and traceability: whether you need leak rate data for each part, serial number traceability and a specific report format.

If you can't set a leak rate limit yet, describe the application instead: operating pressure, coolant type, mounting location and how the part will be serviced. Then ask suppliers for a recommended limit and a quote based on it. You'll get quotes you can compare far more easily than if you leave the field blank or write "zero leak."

How APALTEK Leak Tests Liquid Cooling Components

As a liquid cooling component manufacturer, APALTEK treats helium leak testing as a mandatory step in its quality process. Every cold plate, manifold and liquid cooling module is helium leak tested before it ships, along with pressure testing to confirm sealing.

• Limits set by project: the allowable leak rate, test pressure and any additional pressure hold tests follow each customer's project requirements. If you don't have a spec yet, our engineering team will recommend one based on your operating pressure, coolant and application.
• Assembly testing: we can pre-install the QD brand and model your platform specifies on cold plates or manifolds, helium leak test the complete assembly and ship it as one unit. That cuts the leak risk that comes with installing fittings in the field.

For the full scope of our validation work, including thermal performance, flow and pressure drop, and reliability testing, see our testing and validation capabilities. For custom work, visit our cold plate manufacturer page, or browse our cold plates and manifolds.

FAQ

What is an acceptable helium leak rate for a cold plate?

No single value fits every cold plate. Set the limit based on the consequences of a leak, the operating pressure and the coolant, and state the test pressure and method alongside it. A useful published reference is the OCP quick disconnect specification's 1×10⁻⁵ mbar·L/s before and after cycling tests, but that's a QD qualification limit, not a cold plate production acceptance value.

What's the difference between helium leak testing and pressure decay testing?

Pressure decay testing measures a drop in air pressure inside the part. It's fast and inexpensive but loses sensitivity on large internal volumes and when temperature changes. Helium leak testing measures the tracer gas itself with a mass spectrometer, so it can find much smaller leaks. Many manufacturers use pressure decay to screen for gross leaks and helium to qualify fine leaks.

What's the difference between leak testing and leak detection?

Leak testing happens at the factory, before shipment, to verify that a part is sealed. Leak detection happens in operation: sensors catch a leak in real time and can trigger valves to isolate it. You need both, but they solve problems at different stages. For designing in-service leak detection, see Step 7 of our data center liquid cooling design guide.

Should I choose sniffer or vacuum chamber helium testing?

Choose vacuum chamber testing when you need a quantified, whole-part result in production. Choose sniffing when the part is too large for a chamber or you need to find exactly where a leak is. Many shops use both: the chamber decides pass or fail, and the sniffer locates leaks on rejected parts.

If a part passes a helium test, can it still leak coolant?

A pass means the leak rate was below the limit under the test conditions. If the limit is set sensibly and the test pressure covers the operating pressure, the risk of coolant leakage is low. But proof testing, fitting installation or shipping shock afterward can create new leak paths, which is why assembly-level testing and test sequence matter just as much.

Can helium testing damage or contaminate a part?

No. Helium is inert and won't react with metals or common seal materials, and the test itself is nondestructive. What matters is handling before and after: the part has to be dry before testing, and afterward it should be kept clean and capped so nothing gets into the flow channels.

Put Your Leak Test Requirements in the First RFQ

The earlier your leak test requirements are defined, the more accurate your quotes and the fewer surprises at sample inspection. Send APALTEK's engineering team your drawing, operating pressure, coolant type and target leak rate. If you haven't set a leak rate yet, just describe the application. We'll recommend a test method and acceptance limit, and quote samples to that leak test spec.