How to repaste your GPU for lower hotspot temperatures
Repasting a GPU means removing its cooler, cleaning old thermal compound from the die and heatsink, and applying a fresh interface layer. The aim is to improve heat transfer so the GPU can hold higher boost clocks with less fan noise. It can be worthwhile on an ageing card whose hotspot sensor climbs under load while the rest of the system remains healthy.
The hotspot reading is not the same as the average GPU temperature. It comes from a sensor near the hottest part of the chip package, and the exact sensor layout varies by chip. Some ageing cards gain 5–15°C of hotspot headroom after a good repaste, while a healthy card may gain little. Fresh paste cannot repair a blocked heatsink or failing fan, and ambient temperature can make the same card produce very different results.
Before opening anything, establish that temperature is the actual fault. Document temperatures, clocks, fan behaviour and any artefacts under a repeatable load. If the card is still covered by a consumer guarantee or manufacturer warranty, check whether a managed replacement or authorised repair is available. A patient bench setup and the right materials matter more than speed.
| Likely cause | Best first response | Expected result | Typical effort |
|---|---|---|---|
| Dust or blocked airflow | Clean filters and heatsink fins | Little direct change | 20–60 minutes |
| Failing or noisy fan | Service or replace the fan | Little direct change | 30–90 minutes |
| Dry thermal compound | Clean and repaste the die | Often several degrees lower | 1–3 hours |
| Damaged VRAM pads | Replace pads of matching size | Better memory temperatures | 1–3 hours |
Diagnose the thermal fault before opening the card
Use GPU-Z or HWiNFO64, together with your card’s supported monitoring tools, to record idle temperatures and a sustained gaming or 3D workload. Run each test for at least 15 minutes, keep the room temperature steady, and note the ambient reading. Compare core and hotspot values using the sensors your card actually exposes.
The discipline behind our ROG Delta S review is useful here: realistic sessions and repeatable measurements reveal more than a quick specification check. A GPU should be tested with the same workload, fan state and ambient conditions before and after service. Watch for clock throttling, unstable frame rates, excessive fan noise or visual artefacts.
If the card is clean and the hotspot rises quickly under load, suspect degraded compound. If it is noisy but thermally stable, replace the fan instead. If memory temperatures are excessive, inspect the VRAM thermal pads. Power-limit tuning and undervolving, where supported, can also reduce heat without opening the card.
Gather tools and protect your warranty
Choose a high-quality, electrically non-conductive compound suitable for direct die-and-cooler contact. Follow its application instructions: a thick blob can insulate, while bare spots create voids. For another explanation of thermal interface materials, this thermal interface primer offers a supplementary reading. Avoid liquid metal for a first service unless you already know how to handle bare dies safely.
Work on a stable, non-carpeted bench with good lighting. You need 90% or higher isopropyl alcohol, lint-free swabs, a plastic spudger, a microfibre cloth and the correct screwdriver. You may also need replacement thermal pads, a thermal camera or an infrared thermometer.
Use this preparation checklist before opening the card:
- Photograph the card and group screws by location before removing anything.
- Record each thermal pad’s position, thickness and condition.
- Keep the original compound, shims and cooler parts together in labelled bags.
- Fit an antistatic wrist strap and connect it to a proper grounding point.
- Use compressed air carefully outdoors, without spinning fans or forcing dust into the card.
- Allow enough uninterrupted time for cleaning, mounting and a full load test.
Before buying replacement pads, confirm the exact cooler design and component layout. Generic pads cut from an oversized sheet can be too stiff, too thick or placed in the wrong position.
Remove the cooler without stressing the PCB
Unplug the PC, switch off the PSU and discharge it as its instructions require. Disconnect the PCIe power cables at the graphics card, then release the rear bracket. Photograph cable routing and screw positions, since bracket and cooler screws can differ in length. Keep each group separate.
Lift the cooler straight up, rocking gently if it catches. Thermal compound can hold the cold plate against the die, and some factory pads stick to the cooler. Never lever against the PCB or use a metal tool around exposed components. If a pad lifts, note its orientation and thickness; if it tears, replacement is safer than reusing fragments.
A GPU PCB is heavier and more awkward than it looks. Support the card with both hands, keep fingers away from the cooler’s sharp edges, and avoid pulling a connector sideways. If the board flexes noticeably, stop and reassess the method rather than forcing the cooler free.
Clean the die and apply fresh thermal compound
Remove old compound with a lint-free swab and a small amount of 90%+ IPA. Work in short strokes, then let the metal and die dry completely. Do not flood the area, scrape the silicon or touch the die with bare fingers. Leave any factory shims in place unless the manufacturer’s service documentation explicitly says otherwise.
Apply a thin, even layer of the selected compound to the contact area, following the product’s instructions. The paste should fill microscopic gaps without creating a thick insulating blanket. Use clean applicators for each surface, and do not spread compound over areas occupied by thermal pads. A smeared or uneven layer is a common reason for disappointing repaste results.
Thermal pads require a different approach. Their thickness determines contact pressure, so a pad that is too thick can prevent the cooler from seating, while one that is too thin may not cover the component. Replace damaged or missing pads with matching material and dimensions, and never assume a factory pad can be reused.
Check that the cooler’s contact plate is clean and free of compound ridges. A raised edge or uneven surface can cause the same hotspot problem you intended to fix, so inspect it before applying the new compound.
Reassemble the card and verify the improvement
Reinstall any factory shims and pads in their original positions. Lower the cooler without sliding it across the die, then secure it evenly in a cross pattern. Do not overtighten the screws: excessive pressure can bow the PCB and damage the package. Reconnect the fan and PCIe power cables, and confirm no screw or tool is left inside the case.
Restore the card to the same test position and repeat the original workload for 20–30 minutes. Compare hotspot temperature, sustained clocks, power draw, fan noise and frame-time stability with the baseline. A later test can be more useful than the first, so repeat the check after a day of normal use.
A successful service should show a repeatable improvement under the same conditions. It does not have to produce the lowest possible reading, and a high hotspot sensor reading does not mean the entire GPU is operating at that temperature. Stable performance with lower fan demand is the practical goal.
Account for Australian conditions and consumer rights
Australian Consumer Law provides consumer guarantees for goods sold by Australian businesses, and the ACCC explains that faulty products may need repair, replacement or a refund. A manufacturer warranty is separate, and opening the card can complicate a claim if physical damage occurs. If a fault is covered, use the retailer or authorised service network before attempting a repaste.
International warranty returns can be slow and expensive. Keep the invoice, serial number and case reference, and use tracked shipping through Australia Post or a suitably insured courier. Confirm who pays return freight and what packaging the service centre requires. Do not post a valuable graphics card without checking the carrier’s damage and loss conditions.
Compare Australian prices in AUD, including the 10% GST normally included in local retail pricing. Low-cost overseas listings may gain GST, customs duty and courier charges at delivery. Those additional costs can erase the saving on a replacement card, while international warranty handling may remain impractical.
Local weather shapes the test. During a hot Sydney or Melbourne summer, indoor temperatures can climb above 25°C even overnight; Brisbane adds humidity, while Perth’s dry conditions encourage dust in heatsink fins. Australian mains supplies operate at 230 V, so compatible PC equipment needs no voltage converter, though a surge-protected outlet remains sensible.
Know when to stop and get back to gaming
Stop if the cooler, PCB or die shows damage, or if the card will not seat correctly. If a repaste changes hotspot temperatures very little, investigate mounting pressure, cooler contact, airflow and power limits before applying more compound. Lower core temperatures without better hotspot readings suggest the hotspot interface may still need attention.
Do not repaste a card that is already unstable unless you are prepared to troubleshoot it afterward. Visual artefacts can come from memory, power delivery or drivers, and a fan fault can resemble thermal trouble. Compare tests, change one variable at a time, and keep the original cooler assembly intact until the repair has been verified.
For a cooler, quieter GPU, record a baseline, choose the repair your symptoms justify, and open the card only when you have the right compound, tools and time to reassemble it properly. A clean thermal path gives your GPU more headroom for the next long gaming session.