I have a Gigabyte GeForce RTX 4090 Gaming OC purchased in December 2022. During demanding games, the GPU runs at about 73°C with roughly 100% utilization and 400 W power draw, while the memory reaches around 68°C. The GPU hotspot, however, regularly climbs to 112°C—nearly 40°C above the reported GPU temperature.
Every few minutes, the fans suddenly ramp up to approximately 3466 RPM, creating a lot of noise. The large temperature difference remains even when the fans are not ramping up. I have already tried changing the fan curves in Gigabyte Control Center and MSI Afterburner, updating the NVIDIA drivers and GPU BIOS, and reseating the card.
The card is no longer covered by warranty. Could this hotspot temperature and unusually large delta indicate dried-out thermal paste, poor cooler mounting pressure, or worn thermal pads? Would repasting be the appropriate fix, or should I first try undervolting, reducing the power limit, or lowering the frame rate?
4 Answers
A nearly 40°C difference is unusually large. The most likely causes are degraded or unevenly applied thermal paste, poor contact between the cooler and the GPU die, or a problem with the mounting pressure. Repasting would be the usual next step, but be careful with this card’s thermal pads and putty. If you do it yourself, use a teardown guide specific to the Gigabyte Gaming OC model, since the cooler layout matters.
The GPU temperature and memory temperature look acceptable by themselves, but the hotspot reaching 112°C and triggering sudden fan bursts points to a localized hotspot rather than a general airflow problem. Avoid overclocking while troubleshooting, and monitor whether the card is throttling or showing instability. If you are uncomfortable disassembling a large RTX 4090, professional service is safer than risking damage to the PCB or cooler.
Until you decide whether to open the card, undervolting is a good temporary workaround. Try lowering the power limit to around 70–80%, reducing the core clock slightly, and setting a reasonable frame-rate limit. That should reduce power consumption and fan spikes, though it will not fix the underlying contact or paste issue. A 112°C hotspot should not be ignored if it happens repeatedly.
I would repaste it or have a qualified repair shop do the work. Similar cards with a healthy cooler often show a hotspot delta closer to 10–20°C under load, so a persistent 40°C gap strongly suggests poor thermal transfer. Also inspect the cooler mounting and make sure the replacement materials have the correct thickness; using the wrong pad thickness can make GPU die contact worse.
Those changes helped on my card, but they only reduced the symptoms. After replacing the core interface material and using the correct VRAM putty, the hotspot delta dropped to roughly 10°C.