The issue of diminishing Pentium 4 CPU performance at overheating has been widely discussed (see, for example Intel Pentium 4 1.6A GHz Overclocking, Thermal Monitor). Still, there are all sorts of myths about the Thermal Monitor and Thermal Throttling technology. That is why we decided to refresh our practical knowledge in this field (dropping the theoretical aspect which is abundant in the above-mentioned articles). For this purpose, we conducted a very simple experiment on two top Pentium 4 3.2 GHz CPUs, one with a Northwood core and the other with a Prescott. With the CPU loaded at 100 percent, we powered off the cooler fan and watched the temperature rise and the performance decrease as time went by. This article gives an account of the experiment and its results. To reach a full load of the CPU (or to be exact, of two virtual CPUs, as Pentiums 4 we took support the Hyper-Threading technology), we used a CPU Stability Test from the latest version of CPU RightMark. This enabled us to see the dynamics of CPU performance. It should be noted here that as Thermal Throttling doesn't change the Pentium 4 core frequency, it is performance that we should focus on. As for temperature measurements, we used a quite popular Motherboard Monitor program version 5.3.6.0. Its window (or "Dashboard", to quote the authors) was placed beside CPU Stability Test window (or even overlapped it).

Thermal Throttling in Pentium 4 CPUs with Northwood and Prescott cores

Thanks to Dark_Biene