This one has been debated online since the days of Arctic Silver 5 and it still hasn’t died. So let’s kill it properly.
People are spending $30–$50 on thermal paste believing it’s the difference between a hot system and a cool one. Influencers push Thermal Grizzly Kryonaut like it’s sacred. The community argues about dot methods vs spread methods like the pattern is what matters.
None of it is the problem. None of it is the solution.
The cooler is doing 95% of the work. The paste is filling microscopic air gaps between two metal surfaces. That’s its entire job. Understanding that one fact changes every buying decision you’ll ever make on thermal interface material.
What Thermal Paste Actually Does
Your CPU or GPU die and the base of your cooler are both machined metal surfaces. Under a microscope they are not flat. They have tiny peaks and valleys — microscopic surface irregularities that, when pressed together, leave air pockets between them.
Air is a terrible conductor of heat. Those pockets are insulation you don’t want.
Thermal paste fills those gaps. It displaces the air and replaces it with a material that conducts heat. The cooler then moves that heat away.
That’s it. That’s the whole job. The paste is not cooling anything. The paste is a bridge from the chip surface to the cooler base so the cooler can do its job without air pockets in the way.
The cooler is the variable that matters. The paste is the gap filler.
A premium cooler with cheap paste will always outperform a cheap cooler with $40 paste. Every time. No exceptions.
The Burnout Problem — Why Premium Paste Isn’t Always Better Long-Term
Here’s the part the marketing doesn’t tell you.
Very high-performance thermal pastes are formulated to be extremely fluid for easy spreading and excellent initial contact. That low viscosity comes from a high proportion of low-molecular oils or silicone carriers in the binder.
Under real-world thermal cycling — your system heating up, cooling down, heating up, cooling down, every single day — those carriers migrate. They work their way out from under the contact area. The solid conductive particles left behind become unevenly distributed. Contact thins out. Hotspots form. Thermal performance degrades faster than it would with a more stable, thicker compound.
High viscosity premium paste = better initial numbers on a benchmark, faster degradation in daily use.
A quality silicone-based paste with a stable carrier — mid-range price, nothing exotic — maintains its thermal conductivity longer before degrading. The benchmarks don’t show this because they’re measured at installation. Longevity testing over 12–24 months of real use tells a different story.
The lab test result that keeps coming up: a decent silicone-grease base compound, applied correctly, maintained stable thermal performance longer before burnout than high-end low-viscosity compounds that posted better day-one numbers.
Lower initial temperature in a benchmark is not the same as better cooling over the life of the build. You want the latter.
The Real Numbers
The actual temperature difference between a budget paste and a premium paste on the same cooler is 1–5°C in real-world testing. On a system that’s running 20–30°C under thermal limit under gaming load, that difference is invisible to your performance and invisible to your hardware’s lifespan.
The difference between a Ryzen 5600 on a budget tower cooler with cheap paste versus the same CPU on a 240mm AIO with expensive paste is 15–25°C. The cooler is doing that work. The paste is doing 3°C of it.
Unless you are running a 120W+ TDP chip under sustained overclock — pushing a 13900K or a Threadripper to its absolute ceiling — premium paste is buying you statistical noise. The money is better spent on a better cooler.
What You Actually Need — CPU
A quality mid-range paste. Arctic MX-6 is widely available in Australia, sub-$15, non-conductive, stable, and well-tested. Noctua NT-H2 ships with most of their coolers and is perfectly suited to desktop use. Generic silicone-base compound from reputable brands — the ones that aren’t suspiciously cheap and come with zero documentation — work absolutely fine for any desktop build not running exotic overclocks.
Application method — this matters and most guides get it wrong.
Spread it. Thin, even, full coverage across the entire die surface. Think buttered toast — not a thick glob, a controlled uniform layer that covers the whole surface with nothing left bare.
The “pea in the centre and let the cooler spread it” method gets recommended everywhere as the beginner-safe option because it’s hard to completely destroy. It is not the correct method. Relying on mounting pressure to spread paste evenly across the die assumes the pressure distributes perfectly in every direction. It does not. You get uneven coverage. Thicker in the centre where you started, thinner or absent toward the edges. Air gaps remain where the paste didn’t reach.
The spread method eliminates that entirely. Full die coverage. Controlled layer thickness — thin enough to conduct efficiently, thick enough to fill surface irregularities. No gaps. No guesswork about whether the pressure did the job.
Use a flat spreader — most paste tubes include one, or a clean flat implement will do. Apply a small amount to the die first, then spread it edge to edge in smooth even strokes. The layer should be thin enough that the die surface is just barely visible through it, not buried under a thick white coating.
That’s it. Done correctly, you will not improve on this result with any other method.
GPU Re-Paste — This Is Different and Most Guides Get It Wrong
When you’re re-pasting a GPU — either because thermals have degraded on an older card or you’ve bought used and want to start fresh — the process is not the same as a CPU re-paste. There are multiple components on a GPU that need thermal interface material, and they each need different material.
The GPU Die — Paste
The main GPU processor die gets paste. Same logic as a CPU. A pea-sized amount, centred, spreads under heatsink pressure. Arctic MX-6, NT-H2, or equivalent. Same rules apply.
VRAM Chips — Thermal Pads or Thermal Putty
The VRAM chips are not the die. They are separate memory chips arranged around the die on the PCB. They require thermal pads — not paste. Paste won’t bridge the gap between the chip surface and the heatsink contact point correctly because the gap varies and paste can’t hold a consistent thickness across multiple components.
This is where most people re-pasting their GPU go wrong. They replace the die paste correctly and leave the dried-out factory pads on the VRAM chips. Temperatures improve slightly but not as much as they should. The VRAM is still thermal throttling or running hot because the pads were never addressed.
Thermal putty is a better solution than pre-cut pads for many GPU applications. It’s a phase-change material — solid at room temperature, softens under heat, conforms exactly to the gap between each individual VRAM chip and the heatsink plate above it. No thickness mismatch. No air gaps. Better conformity than a rigid pad across components that aren’t all exactly the same height.
Quality thermal putty is available from AliExpress for a fraction of the cost of branded options. The phase-change property means it lasts significantly longer than both paste and conventional pads before needing attention again.
VRM Components — Thermal Pads
The VRMs (voltage regulation components) around the GPU die also need thermal pads to contact the heatsink backplate or frame. Same rule — measure the gap, use the correct thickness pad. Don’t leave these dry.
The Re-Paste Shopping List for a GPU
-
GPU die: Quality paste (MX-6, NT-H2 equivalent)
-
VRAM chips: Thermal putty (AliExpress, phase-change) — enough to cover all chips
-
VRM components: Thermal pads, correct thickness for your specific card (check a teardown guide for your card model)
-
IPA 90%+: For cleaning old compound and pad residue off all surfaces before re-applying
Thermal Capacity vs Cooling — The Concept Worth Understanding
Thermal capacity is how much heat a material can absorb before its temperature rises. Thermal conductivity is how quickly it transfers that heat onward.
Your cooler has high thermal capacity (the heatsink mass absorbs and distributes heat) and high thermal conductivity (copper/aluminium moves heat fast). The paste has neither — it’s a thin layer with the singular job of bridging the gap.
Chasing thermal conductivity numbers on paste marketing (W/mK ratings) misses this entirely. The W/mK of the paste matters at the margins when the layer is correct and the cooler is capable. Buying paste with a 12 W/mK rating instead of 8 W/mK and expecting the system to suddenly run cooler is buying marketing, not results.
Buy more cooler. Apply decent paste correctly. Move on.
The TrojanHQ Standard
For any desktop build:
-
Mid-range non-conductive paste — Arctic MX-6, Noctua NT-H2, equivalent
-
Pea-sized application, centred
-
Replace every 3–4 years on a desktop, sooner if thermals degrade
-
Don’t use liquid metal unless you are an experienced builder running a specific high-performance application and you understand the risks
For GPU re-paste:
-
Paste on the die
-
Thermal putty on VRAM chips
-
Correct thickness thermal pads on VRM components
-
IPA clean everything first
The argument about which premium paste brand is 2°C better than another brand is a discussion for people who have already maxed out every other variable in their thermal chain. If you haven’t done that — and almost nobody has — it’s not the discussion you need to be having.
King Frost — TrojanHQ