My RTX 3070 has recently started overheating during games, especially when I uncap the frame rate. It usually reaches 80–85°C with a hotspot around 100–105°C, and the hotspot temperature drops almost instantly to about 65°C when I tab out of a game. Eventually the computer black-screens, and I cannot recover it with the usual graphics-driver restart shortcut. I previously had poor case airflow, but I fixed that and also replaced the CPU cooler after finding its thermal paste had dried out. The GPU is about five years old and worked reliably until I began playing a more demanding game. I have already considered undervolting or underclocking, but those feel like workarounds rather than fixing the underlying issue. Would repasting the GPU be worthwhile, and should I replace the thermal pads or use thermal putty as well?
3 Answers
Those core temperatures are not automatically abnormal for an RTX 3070, but the 20–25°C gap between the core and hotspot, loud fans, and eventual black screens suggest the cooling system deserves investigation. Try a sensible undervolt and a frame-rate cap first because they are low-risk and can greatly reduce heat with little performance loss. Opening the card should generally be a later step, since an uneven remount or incorrect pad thickness can make temperatures worse.
Repasting can help, but thermal pads are more complicated because their thickness affects cooler contact. Do not replace them with random pads or putty unless you know the correct application and thickness for your exact card. Thermal putty can work for some designs, but it is not automatically better than the original pads. Undervolting is still useful after the repair because it reduces power and heat rather than merely masking a fault.
A repaste and undervolt together would be the most practical approach. If temperatures remain high afterward, check for a failing fan, clogged heatsink, poor cooler contact, or another hardware problem before continuing to push the card.
At five years old, repasting is a reasonable next step, especially if the card used to run cooler and now immediately drives the fans to full speed. Use a high-quality paste, clean the old material carefully, and tighten the cooler evenly in a cross pattern. Also inspect the fans, heatsink fins, and mounting pressure before assuming the paste is the only problem.
I had already tried undervolting with a custom fan curve and it did not make much difference. Before a crash, the fans ramp up dramatically, so I am increasingly suspicious that the paste has dried out after years of heat and previously poor airflow.