My Radeon RX 6800 XT overheats while gaming, loses connection, and often crashes with a GPU timeout error—even in relatively undemanding games during loading. A stress test reaches a 110°C hotspot temperature in about five seconds. I've reinstalled the drivers, reseated the card, and checked the case airflow, but none of that helped. The card is only reliable when I underclock and undervolt it while using an aggressive fan curve. I'm worried the card may be failing, but I can't afford to replace it. Would it be reasonable to keep using it with conservative settings, or should I pay to have the thermal paste or other components serviced before the problem gets worse?
2 Answers
Repasting the GPU die is a sensible alternative to replacing the card. Follow a teardown guide for the exact model and be careful with the cooler, screws, cables, and thermal pads. A phase-change material designed for GPU dies can also be worth considering, since it can perform well over repeated heating and cooling cycles. Until then, conservative clocks, undervolting, and a stronger fan curve are reasonable temporary measures, but they’re treating the symptom rather than fixing the underlying cooling problem.
The symptoms strongly suggest the thermal interface on the GPU has degraded, especially since the hotspot reaches 110°C almost immediately. Replacing the paste is probably the most practical repair and shouldn’t cost hundreds if you do it yourself. Thermal pads may also need replacing, but make sure you find the exact thicknesses for your specific card before ordering them—incorrect pad sizes can make cooling worse.
Thanks! I was concerned that opening the card could damage it, but if the process is reasonably straightforward I may try it carefully rather than pay for an expensive repair.