I bought a custom-built gaming PC and recently noticed that the RTX 5090 was running unusually hot. I returned it for inspection, and the technician said a cable had come loose. They reported that the GPU stayed around 57°C during their full-load test, but after bringing the computer home, it reached about 85–87°C while gaming, with a hotspot around 93–96°C.
The PC was transported carefully, so I'm wondering whether another cable could have loosened during the trip or whether the liquid-cooling system is malfunctioning. The system is sitting on a hardwood floor with plenty of clearance around it. HWiNFO also shows one GPU fan at roughly 3,287 RPM while the second reports 0 RPM, even though both are commanded to run at 100%. More importantly, I noticed that the three fans mounted to the GPU radiator are not spinning.
The system includes an ASUS ROG Astral RTX 5090 LC OC, a Lian Li O11 Dynamic EVO XL case, an AMD Ryzen 7 9800X3D, an Arctic Liquid Freezer III Pro 360 radiator, and a 1200W power supply. I'm not very experienced with PC hardware and would prefer not to open anything or risk damaging the components. What simple checks can I perform, and should I return it to the builder?
3 Answers
Radiator fans that never spin under gaming load need to be investigated. With the PC powered off, you can visually check whether the fan cables are connected to the appropriate motherboard or controller headers, but don’t force any connectors or disassemble the GPU cooler. Since the machine was professionally assembled and the fan problem is clear, the safest option is to return it and specifically point out that the radiator fans remain stopped and that one GPU fan sensor reads zero.
The core-to-hotspot difference is only about seven degrees, which generally suggests the cooler is making decent contact with the GPU. That makes a completely failed cold plate or badly mounted block less likely. The more suspicious detail is one fan reporting 0 RPM while being commanded to run at full speed, along with the radiator fans not moving.
The shop’s 57°C result may also have been measured on an open test bench rather than inside the closed case. You can remove the side panel and run the same game for around ten minutes as a simple diagnostic. If temperatures drop substantially, case airflow is contributing. If they barely change, stop testing and have the builder inspect the fan and pump wiring.
First check the physical airflow and the fans themselves. Keep the case on a solid surface, leave a few inches of clearance around the intake and exhaust areas, and make sure nothing is blocking the top. Also look directly at the GPU fans while the system is under load. If both GPU fans are visibly spinning, the 0 RPM sensor may simply be inaccurate. If one is stopped, the builder should repair or replace the card.
However, if all three fans on the GPU radiator are actually stopped, that is not normal. A disconnected fan or pump header could explain the temperature, so I’d have the shop check the connections rather than opening the cooling system yourself.
The computer is on hardwood with about a foot of clearance on each side. I confirmed that the three radiator fans are not spinning. Could this be another loose cable, even though the PC is only about six months old?
The liquid cooler itself appears to be working, but the three radiator fans are still stationary. Since this is a relatively new system, I’m wondering whether their cable has come loose again.