All three exist to solve the same problem: closing the microscopic gap between a chip and its cooler so heat has somewhere to go. They differ in how they close that gap, how much heat they can move, and how forgiving they are if something goes wrong.
Thermal paste
The default choice for CPUs and GPUs. It's a syringe-applied compound, typically carbon-based, metal-oxide, or silicone-based, that stays semi-liquid and fills surface imperfections under the cooler's mounting pressure. Conductivity for mainstream pastes runs roughly 8-16 W/mK, non-conductive formulas carry no short-circuit risk, and most last multiple years before needing reapplication. It requires opening the system to apply and, eventually, to reapply.
Thermal pads
Solid sheets rather than a spreadable compound, pads are common on GPU memory chips, VRMs, and other components with uneven or multiple contact points rather than a single flat die. They're easier to install correctly since there's no risk of applying too much or too little, but standard silicone-based pads generally have lower conductivity than a good paste and can be more expensive per application. Phase-change pads like Honeywell PTM7950 are a notable exception, solid at room temperature but softening at operating temperature into something that behaves much like a very stable paste, while still avoiding the mess of a syringe.
Liquid metal
Gallium-based compounds like Thermal Grizzly Conductonaut sit in a different performance tier entirely, often 70-80+ W/mK versus 8-16 W/mK for standard paste. That's a real difference for delidded CPUs, direct-die cooling, and flagship chips pushed to their thermal limits. The tradeoff is electrical conductivity: liquid metal will short a board if it contacts the wrong surface, and it reacts with bare aluminum, so it's unsuitable for coolers with aluminum contact plates. It demands more careful application than paste or pads and isn't recommended for a first thermal paste job.
Quick comparison
- Typical conductivity. Thermal paste: 8-16 W/mK. Thermal pad: varies, generally lower for standard pads. Liquid metal: 70-80+ W/mK.
- Electrical safety. Thermal paste: non-conductive in most formulas. Thermal pad: non-conductive. Liquid metal: conductive, with short-circuit risk.
- Application difficulty. Thermal paste: low to moderate. Thermal pad: low. Liquid metal: high.
- Reapplication. Thermal paste: every few years typically. Thermal pad: rarely, or never for phase-change pads. Liquid metal: requires care, but stable long-term.
- Best for. Thermal paste: mainstream to high-end CPU/GPU builds. Thermal pad: GPU memory/VRM and uneven surfaces. Liquid metal: delidded CPUs and extreme overclocking.
Which one should you use
For a standard CPU or GPU cooler installation, thermal paste covers the vast majority of builds safely and affordably. Reach for pads when you're covering GPU memory chips or components a syringe can't apply evenly to, or when you specifically want to avoid ever reapplying, in the case of phase-change pads. Liquid metal is worth the extra care only when you're chasing every last degree on a delidded or extremely high-TDP setup and you understand the compatibility and short-circuit risks going in.
Compare specific products across all three categories, with conductivity, electrical safety, and package details, on the Consumercarts thermal paste page.
Frequently asked questions
Can I use liquid metal on any CPU cooler?
No. Liquid metal reacts with bare aluminum, so it's unsafe on coolers with an aluminum contact plate. Confirm your cooler's contact surface material (copper and nickel-plated surfaces are generally safe) before using a liquid metal compound.
Is a thermal pad ever a direct substitute for paste on a CPU?
Not typically for the main CPU die. Standard pads are usually thicker and lower-conductivity than paste, making them better suited to secondary components like GPU memory rather than the primary CPU-to-cooler contact point. Phase-change pads are the exception built specifically for that primary contact use case.
Does higher conductivity always mean lower temperatures in practice?
Not proportionally. A mainstream CPU running at stock settings may see little real-world difference between an 8 W/mK paste and a 16 W/mK paste. The gap becomes meaningful under sustained high-TDP loads or overclocking, where every degree of headroom matters.