GPU Thermal Pads vs Thermal Paste: When Pads Degrade and What to Replace Them With
Most attention on GPU thermal maintenance focuses on the paste covering the die itself, but the thermal pads covering VRAM and VRM components dry out and lose conductivity too, often silently, on a card that still looks fine from the outside.
Opening up a graphics card for a repaste reveals two distinct thermal interface materials doing different jobs. Thermal paste sits between the GPU die and the cooler's vapor chamber or heat pipe base, chosen for high conductivity in a thin layer since the die is the single hottest, most concentrated heat source on the board. Thermal pads sit between the cooler's baseplate and the surrounding VRAM modules and VRM power stages, chosen for their ability to fill an uneven gap between components sitting at different heights on the PCB, since a rigid paste cannot bridge that height variance the way a compressible pad can.
Pads and paste degrade through different mechanisms and on different timelines. Paste dries out as its liquid carrier evaporates over years, gradually losing conductivity and sometimes cracking or pulling away from the surface. Pads degrade primarily by losing their compressibility and conforming ability over repeated thermal cycling, and cheaper pads used by some board partners on lower-tier cards can be undersized for the actual gap, compressed too much or too little from the factory, well before any age-related degradation even starts.
Recognizing Which One Has Failed
| Symptom | Likely Cause |
|---|---|
| GPU core (hotspot) temperature climbed over time, VRAM/VRM temps stable | Die paste degraded |
| Memory junction/VRAM temperature climbed, core temp stable | VRAM thermal pads degraded or undersized from factory |
| VRM temperature high under sustained load, other sensors normal | VRM thermal pad issue, separate from the memory pads |
| All sensors elevated together | Likely a cooler mounting or fan issue rather than a specific pad or paste failure |
HWiNFO64 exposes separate sensor readings for GPU hotspot, memory junction temperature, and VRM/MOS temperature on most modern cards, making it possible to isolate which specific interface has degraded rather than guessing based on overall card temperature alone.
What to Replace Pads With
- Match the thickness carefully. Pads come in various millimeter thicknesses to fill a specific gap; a pad that is too thin leaves an air gap, while one that is too thick can prevent the cooler baseplate from seating flat against the die, indirectly hurting the paste's contact too.
- Higher conductivity pads (rated in W/mK) are worth the modest price difference over generic low-end pads for VRAM specifically, since GDDR6X and GDDR6 memory on high-end cards run hot enough that pad quality measurably affects sustained memory clock stability under overclocked or heavy workloads.
- VRM pads can generally use a more moderate conductivity rating than VRAM pads, since VRM components tolerate a wider temperature range before performance is affected, but replacing a hardened, cracked pad is still worthwhile whenever the cooler is already off the card for other maintenance.
- Photograph the pad layout and thicknesses before removing the old ones, since aftermarket pad kits sold for specific card models are the easiest way to get correct thickness without trial and error, but a generic multi-thickness pad kit works fine if you record what was removed from where.
A full repaste-and-repad is worth doing together rather than addressing paste and pads separately across two sessions, since the cooler has to come off the card either way and the labor cost of reopening it a second time outweighs the modest cost of doing both while everything is already disassembled.
How Long Before Pads and Paste Actually Need Attention
There is no fixed universal timeline, since degradation speed depends heavily on how hot the card has run over its life and how many thermal cycles it has been through, but a rough guide based on common reports is that cards used for sustained heavy gaming or mining-style continuous load start showing measurable temperature creep from paste degradation somewhere in the three-to-five-year range, while lighter, more intermittent use can push that well past five years before it becomes noticeable. Pads tend to hold up somewhat longer than paste on average, since they are not relying on a liquid carrier that can evaporate, but factory pads on lower-tier cards that were undersized or poorly seated from day one can show problems much earlier, sometimes within the first year or two.
Tracking GPU hotspot temperature over time, rather than relying on a one-off measurement, is the most reliable way to catch gradual degradation before it becomes severe enough to cause throttling or instability. A card that ran comfortably at a given hotspot temperature under a specific benchmark a year ago and now runs measurably hotter under the identical test, with no other configuration changes, is showing exactly the kind of slow creep that a preventive repaste addresses before it becomes a performance problem rather than just a curiosity.