Graphics cards use thermal pads, not just paste, to cool memory chips, VRM components, and sometimes the backplate. Over years of heat cycling those pads dry out, harden, and lose contact, which pushes memory junction temperatures higher and can trigger throttling. Replacing them is a rewarding afternoon job that can drop VRAM temps by 10-20C, but it requires the right pad thickness and a careful hand.
As an Amazon Associate we earn from qualifying purchases at no extra cost to you.
Why pads matter more than people think
GDDR6 memory can safely run warm, but sustained temperatures above 95-100C cause the card to slow its memory clocks to protect itself. On heavily used cards, the original pads are often the weak link long before the main GPU paste is. Fresh, correctly sized pads restore the gap-filling contact between hot chips and the cooler.
Get the thickness right first
This is the step that ruins most first attempts. Pads come in thicknesses from 0.5mm to 3.0mm, and the gap between each component and the cooler differs. Using a pad that is too thin leaves an air gap; too thick and the cooler will not seat, or it warps the PCB. If you cannot find a teardown that lists your card’s pad thicknesses, measure the compressed originals or buy a variety pack.
| Component | Common pad thickness | Typical conductivity |
|---|---|---|
| VRAM (memory chips) | 1.0-2.0mm | 8-14 W/mK |
| VRM / MOSFETs | 1.5-3.0mm | 8-14 W/mK |
| Backplate contact pads | 1.0-2.0mm | 3-8 W/mK |
| Main GPU die | Paste, not pad | 8-14 W/mK |
Step by step
1. Document and disassemble
Photograph every screw location and pad before you touch anything. Remove the backplate, then the cooler screws in a cross pattern. Disconnect the fan header. Note which pad thickness sits where, since they are rarely all the same.
2. Clean everything
Peel off old pads and wipe residue with 90%+ isopropyl alcohol and a lint-free cloth. Clean the GPU die and cooler cold plate for fresh paste. Do not scrape metal contacts with anything sharp.
3. Apply new pads and paste
Cut pads to match each chip’s footprint, peel both protective films, and press them into place. Apply a small amount of quality thermal paste to the main die. Confirm no pad overhangs onto neighboring components where it could short a circuit if it is a conductive type (most are not).
4. Reassemble and test
Tighten cooler screws gradually in a cross pattern for even pressure. Reattach the backplate and fan header. Boot, run a GPU stress test, and watch memory junction temperature in a monitoring tool. A well-done repad usually shows a clear drop within minutes.
Risks to respect
Repadding voids many warranties, so check yours first. Wrong pad thickness is the top cause of higher-than-before temps or a cooler that will not close. Work on a static-safe surface and never force screws.
Frequently asked questions
How often do GPU thermal pads need replacing?
Most cards go 4-6 years before pads dry out enough to matter. If your memory junction temps have crept up 10C+ over a card’s life or you see memory-related throttling, it is a good candidate.
Can I reuse the original pads?
Once removed, pads lose their gap-filling shape and rarely reseat well. Budget for fresh pads of the correct thickness rather than reusing hardened originals.
Bottom line
Replacing GPU thermal pads is a high-value maintenance job when memory temps climb, but success hinges entirely on matching pad thickness and applying even cooler pressure. Photograph everything, buy a thickness variety pack if unsure, and you can reclaim 10-20C on an aging card in an afternoon.
