Your latest 4090 is experiencing thermal throttling issues.
Your latest 4090 is experiencing thermal throttling issues.
I recently switched from a 3070 to an MSI liquid cooled 4090, but it doesn’t maintain the right temperature. It stays cool when idle, around 35-40°C, but it jumps to 90°C under load and affects performance. I’m not sure if the problem is with my setup—my radiator is positioned above the pump, airflow seems normal, and fans are directing air into the radiator. I’ve attached a picture below in case you need to review it. Could this be related to software or BIOS issues, or is the hardware itself damaged? I’m running a 7800x3d on an ASUS Prime X870-P board, so I’m wondering if either of those might be contributing.
Consider adding a second fan set on the opposite side of the GPU radiator for a push/pull setup, which can enhance airflow. Use high static pressure fans to maximize air movement through the radiator. Also, please share images of the back side of the case so we can examine the GPU radiator layout.
What monitoring method are you employing? Are you certain you’re checking the GPU core temperature? If we assume that’s the core itself is overheating, it could mean poor thermal contact or a faulty pump. I’m not well versed with the liquid used by AIO systems to determine if you can track pump RPM through MSI software. Clearly, with this setup we don’t have information about its past usage or any prior issues. If the system reports 90°C temperatures, simply adding more fans won’t fix it. A water-cooled card without radiator fans shouldn’t reach 90°C as long as the pump is functioning.
I’ve noticed these readings in afterburner, icue, and GPU-Z. I don’t think MSI offers software that displays pump RPMs, at least not publicly. I’ve spoken to the person who sold it and he says he rarely sees temps above 60°C. I’m not sure what to do next.
I set up the card with a push/pull fan setup, but it didn’t solve the problem. Right now there aren’t any extra fans installed, and I’ve been working on troubleshooting without expecting a big impact.
Start with simple checks. Is the afterburner detecting pump RPM? Are there any sensor readings in HWiNFO? Did the prior owner remove the cooler from the card? You may need to look at the mounting connection and thermal paste, possibly replacing it.
I’m checking hwinfo but no pump info appears in afterburner. It looks like the previous owner didn’t remove the cooler, possibly downgrading to a 5070 because of GPU size. I’m thinking about inspecting or reapplying thermal paste before replacing the cooler entirely. Thermal Grizzly’s Kryo is designed for GPUs, but I’m not sure if a generic paste works—maybe it needs a specific type.
I performed several stress tests on the sensors while observing hwinfo’s sensors, and the card maintained around 75° at a sustained frequency of 2850mhz. It handled approximately 11,200 in Blender 4.3. I’m really puzzled but I’ll move forward—I think my problem is fixed.