I've been running the Devstral:24b model locally through Ollama, and HWiNFO reports that my Intel Core i5-13600KF is thermal throttling. I'm using a Corsair iCUE H150i 360mm liquid cooler and replaced its original thermal paste with Noctua NT-H2, so I expected lower temperatures. HWiNFO recorded a CPU package temperature of 100°C, peak core clocks around 5,088 MHz, up to 78% total CPU usage, 100.4% CPU utility, and roughly 185W of CPU package power. At idle, the CPU sits around 42–46°C. The coolant reached about 40°C, the radiator fans peaked at 2,100 RPM, and the pump reached roughly 2,689 RPM. The radiator is mounted at the top of the case, and the pump is detected in the BIOS. Is this behavior expected for this kind of workload, or could there be a problem with the CPU, cooler installation, pump, or fan settings?
3 Answers
A 360mm AIO does not guarantee that a modern CPU will stay below 100°C when it is allowed to boost hard. The 13600KF can produce a lot of heat, and an AI model may create a sustained workload that behaves differently from ordinary gaming. Thermal throttling is only a serious concern if it happens during normal lighter use, the clocks collapse badly, or temperatures remain at the limit constantly. For a practical test, run a benchmark with the BIOS power limits enabled, compare core temperatures, and check for a large temperature difference between cores. One unusually hot core can point to mounting or contact problems.
Your coolant temperature is the more useful clue here. A coolant temperature around 40°C suggests the radiator is receiving and removing heat, so the fans and pump may be working normally. If the CPU reaches 100°C while the coolant stays relatively cool, check the cold plate mounting, the amount and spread of thermal paste, and whether the protective film was removed. Uneven contact can cause the CPU to run hot even with a large radiator. Also confirm that the pump is connected to a full-speed pump header or otherwise configured to run at maximum speed.
The missing detail was the CPU model, but with a 13600KF this temperature isn’t automatically surprising. Local AI workloads can keep many CPU cores busy and make the processor draw around 185W. Intel’s boost behavior will use available thermal and power headroom until the chip reaches its temperature limit, so briefly hitting 100°C and throttling can be expected under a sustained heavy load. It’s still worth checking BIOS power limits and whether the motherboard is running an aggressive unlimited-power profile.
That makes sense. I wasn’t aware the motherboard could allow the chip to use more power than the default limits. I’ll check the BIOS settings and compare temperatures with stock power limits.
The radiator is top-mounted, and the pump is visible in the BIOS. I’ll still inspect the mounting pressure and cooler contact to make sure nothing is preventing an even mount.