I bought a used GPU from a private seller. The card works normally, but in one of my systems its fans stay at a normal speed until the GPU experiences even a tiny amount of load, at which point they jump straight to 100% and remain there. This can also happen in the BIOS, and changing motherboard fan-control settings seems to trigger it. A full restart temporarily restores normal fan behavior.
The same GPU works correctly in a second system: its fan speed increases and decreases appropriately with temperature and load. I reinstalled Windows, reset the BIOS to optimized defaults, tried different PCIe slots, changed power cables, and tested various PCIe, power, and fan-control settings.
In manual fan control software, setting the GPU fan to 20% or 40% still makes it run at full speed. The fans are not completely stopped before the problem occurs, and the reported GPU temperature is around 40–50°C. What could make the card behave normally in one system but switch between ordinary speed and 100% in another?
2 Answers
Update the motherboard BIOS, preferably by downloading the correct version to a USB drive and using the board's built-in flashing utility. Also verify that the GPU's 8-pin and 6-pin power connectors are each properly connected and, when possible, use separate PSU cables for them. Inspect and clean the PCIe slot in the problematic system, then make sure the card is fully seated. These steps may rule out a motherboard, PCIe-slot, firmware, or power-delivery compatibility problem.
Check the GPU's sensor readings with a hardware-monitoring tool while the fans ramp up. Look at the core temperature, hotspot, and memory-junction sensors. If one sensor is reporting an impossible value or is stuck at a fixed extreme, the card may be deliberately running the fans at full speed because it thinks there is an overheating condition. This can happen with a used card that has been disassembled or repasted.
Also test manual fan control. If every setting results in either normal/off behavior or full speed, rather than an intermediate speed, the card may not be receiving PWM control correctly. That could indicate an issue with the fan connection or replacement fans that do not support the expected control signal. Since the card works in another computer, comparing sensor readings and fan-control behavior between the two systems should help isolate the cause.
Manual control does produce the binary behavior: even setting the fan to 20% makes it spin at 100%. However, the card works normally in the other Windows system, so I will compare the sensor readings there as well. The core temperature in the problematic system is only around 40–50°C, but I still need to check hotspot and memory-junction temperatures.
Both GPU power connectors are connected separately, and I already tried swapping the cables and checking the connections. I attempted an online BIOS update but did not find one, so I will try updating from a USB drive and inspect the PCIe slot more carefully.