Explained

Thermal Paste, Explained

After reading, you'll understand why thermal conductivity ratings are close to useless for comparing brands, what specs to read instead, and when liquid metal is worth the risk.

Updated August 23, 2026 · Companion to our Tech & Gadgets review

In this guide

The 30-second version
The products
What actually matters
Specs in plain English
Green flags vs red flags
The brands
How to read a listing
How much to spend
FAQ

Almost every thermal paste listing leads with a thermal conductivity figure in W/mK, and almost every buyer treats a bigger number as a better product. That instinct is wrong in a specific and interesting way: the figure is measured differently by different companies, the relevant standard calls its own output apparent thermal conductivity because these materials are not homogeneous, and several major brands have stopped publishing the number entirely. Meanwhile the specs that decide whether a paste damages your hardware or dries out in two years usually sit further down the page.

The 30-second version

The W/mK number is not comparable between brands, because makers use different test methods and several no longer publish one at all.
The real spread between good pastes is small, with Thermal Grizzly's own datasheet showing 0.06 K between its flagship and its standard paste.
Non-conductive and non-capacitive are different claims, and a paste can carry the first without the second, as Arctic Silver 5 does.
Liquid metal and aluminium must never meet, since gallium alloys attack aluminium and manufacturers list it as a substance to be avoided.
Pump-out and dry-out are what actually kill a joint, so a stated service life is worth far more than a headline conductivity figure.

The products this guide is about

Our verdict on each, with the full reasoning on the companion review.

What actually matters

Electrical conductivity - decides whether a stray smear near pins is a wipe-up job or a dead motherboard. Filled greases are almost always non-conductive; liquid metals are conductive by definition and must be treated as live.
Capacitance - a paste can fail to conduct current and still store enough charge to misbehave if it bridges two close traces. Arctic Silver flags this for Arctic Silver 5, while Arctic explicitly rules it out for MX-6.
Metal compatibility - normal pastes work on any cooler base, but gallium-based liquid metals attack aluminium and amalgamate with copper, silver and gold, so they want nickel plating.
Bond line thickness - the thinner the layer of paste in the finished joint, the less the paste's own conductivity matters. Good mounting pressure and a flat cooler base shrink the bond line more than upgrading paste does.
Pump-out resistance - repeated heating and cooling makes the joint breathe and can squeeze paste sideways out of the middle. Viscosity is the main defence, which is why some good pastes are deliberately stiff.
Burn-in or cure behaviour - most modern pastes reach full performance instantly, but a few take days of thermal cycling to settle. If you benchmark straight after mounting, know which kind you have.
Stated service life - very few makers commit to how long the paste should stay good on a running CPU. Where one does, it is a more meaningful quality signal than any performance claim.

The specs, in plain English

Thermal conductivity (W/mK)How well a material moves heat through its own bulk, in watts per metre-kelvin. It is the number on the box, and it is a poor guide here because pastes are mixtures rather than uniform materials and brands measure it differently. Arctic says realistic values are 1 to 4 W/mK and that anything far above that is not credible.

Thermal impedance or thermal resistanceHow much the whole joint resists heat flow, usually quoted in K/W or degrees per watt. This is closer to what you feel, because it includes the contact between paste and metal rather than just the paste itself. Thermal Grizzly quotes 0.0032 K/W for Kryonaut instead of a conductivity figure.

ASTM D5470The usual test method for thermal interface materials, which measures steady-state thermal impedance in an idealised setup. The standard itself says the impedances it measures cannot be applied directly to most real applications, because real joints are not uniform, parallel heat paths. It also uses the phrase apparent thermal conductivity on purpose.

Bond line thickness (BLT)The thickness of the paste layer once the cooler is bolted down, typically measured in microns. Thinner is better, and the thinner it gets the less the paste's own properties matter. This is why liquid metal wins: it forms a far thinner layer than particle-filled grease can.

Pump-outThe gradual migration of paste out of the centre of the joint as the chip and cooler expand and contract at different rates. It leaves a starved patch right over the hottest area, so temperatures creep up over months. Stiffer, higher-viscosity pastes resist it better.

Dry-outThe carrier oil evaporating or bleeding away over time, leaving dry filler that no longer conforms to the surfaces. It is why old paste turns crumbly and grey. Makers describe resistance to it with phrases like non-drying and non-bleeding rather than with a test figure.

Volume resistivity and breakdown voltageTwo hard numbers that describe how well a paste refuses to pass current. Arctic quotes 1.8 x 10^12 Ohm-cm and 7.5 kV/mm for MX-6, which is far more useful than a vague safe to use claim. If a maker publishes these, take it as a sign they measure things.

Eutectic alloyA metal mixture whose melting point is lower than any of its ingredients. Liquid metal thermal compounds use a gallium, indium and tin eutectic so the alloy stays liquid at ordinary operating temperatures. Below about 8 C it stops behaving, which is why makers tell you to warm it before applying.

Phase change material (PCM)A thermal interface that ships as a solid pad and melts into a thin film the first time the chip gets hot. It is a third family alongside grease and liquid metal, and ASTM D5470 covers it explicitly. The upside is a very consistent bond line; the downside is you cannot adjust it once it has flowed.

Green flags vs red flags

Green flags

The maker publishes electrical specs such as volume resistivity or breakdown voltage, not just adjectives.
A stated operating temperature range with both ends given, rather than a single maximum.
A resealable syringe and a stated storage life, which shows the maker expects you to keep it.
An explicit statement about whether burn-in is required, either way.
A named service life or usage period on the CPU rather than a phrase like long-term stability.

Red flags

A very large W/mK figure with no test standard named anywhere on the page or datasheet.
Marketing that leans on exotic ingredients such as diamond or silver content instead of measurable properties.
No statement at all about electrical conductivity, which is the one safety question every listing should answer.
Liquid metal sold without aluminium warnings, application tools or a safety data sheet.
No batch, revision or authenticity information from a brand whose products are widely counterfeited.

Who's who: the brands

Noctua - Austrian cooling specialist whose pastes are conservative, non-conductive and unusually well documented on storage and service life.
Arctic - High-volume German brand that competes on price-to-performance and publicly refuses to quote thermal conductivity figures at all.
Thermal Grizzly - German enthusiast brand covering everything from budget greases to liquid metal, with full datasheets and safety data sheets for each.
Arctic Silver - Long-established US maker of silver-loaded compounds, best known for Arctic Silver 5 and its lengthy break-in period.
Cooler Master - Broad-line component maker whose current CryoFuze pastes list viscosity and specific gravity but, like the others, no conductivity figure.

How to read a listing without getting fooled

Start at the bottom of the listing, not the top. Skip the headline conductivity claim and find the technical table: you want an operating temperature range with both ends stated, a density and viscosity figure, and something concrete about electrical behaviour. If the maker gives volume resistivity and breakdown voltage, that is a good sign they measure their own product. Next, look for the words burn-in, cure or break-in, because a paste that needs 200 hours of thermal cycling behaves very differently from one that is at full performance the moment you tighten the bracket. Then check what the maker says about metals, which only matters if you are looking at liquid metal but matters absolutely in that case. Finally, look for anything resembling a durability commitment. Almost everyone says non-drying and long-term stable; almost nobody puts a number on it, and the one who does is telling you something.

How much should you spend?

This is a category where spending more buys you very little performance and occasionally buys you real peace of mind. Budget tier pastes from reputable makers already give you non-conductive filler, a resealable syringe and no burn-in, and they are the right answer for most builds; the value here is genuinely high because the difference between a competent paste and an excellent one is a degree or two. Mid-range tier gets you better documentation, longer stated service life, sometimes a higher temperature ceiling, and cleaning wipes in the box. Premium tier is mostly liquid metal and specialist compounds, where you are paying for a different physical mechanism rather than a better version of the same one, and where the risk profile changes completely. A useful rule: if the price gap between two pastes would cover a meaningfully better cooler, buy the cooler.

Questions, answered

Does more expensive paste run cooler?

Barely, and there is manufacturer data to show it. Thermal Grizzly's Kryonaut datasheet charts its own range on an identical test and the whole line spans about 2.6 degrees, with only 0.06 degrees between the flagship and the standard paste. A competent budget paste from a reputable maker gets you almost all of the available benefit. Spend the difference on cooler surface area or fan quality instead.

How often should I replace thermal paste?

Less often than the internet suggests, though it depends on the paste and the thermal load. Noctua publishes a recommended usage time of up to five years on the CPU for NT-H2, which is the clearest guidance any maker in this group gives. If your temperatures have crept up over a year or two under the same workload, that is a symptom of pump-out or dry-out and a re-paste is worth doing. Otherwise, leave it alone.

Is silver paste better than ceramic or metal oxide paste?

Not in any way you will notice, and the filler type is largely a marketing distinction now. Silver-loaded compounds were genuinely ahead when they arrived, but modern metal oxide and ceramic formulations perform comparably in a thin bond line. Silver content also brings a downside, since silver-filled compounds tend to carry capacitance warnings that non-metallic ones do not. Judge on the published electrical and durability specs, not on the ingredient.

What happens if I use too much paste?

Usually nothing dramatic, because the excess squeezes out at the edges when you tighten the cooler. The real risk is what the squeeze-out lands on: a non-conductive paste is a cleanup job, while liquid metal on a socket or on surface-mount components can short things permanently. Too little paste is the worse mistake, since gaps trap air and air is a terrible conductor. Aim for full coverage under mounting pressure and wipe up anything that escapes.

Ready to choose? See our top pick, the alternatives, and the one we'd skip See the Thermal Paste verdicts