CPUID HWMonitor uses the incorrect TJMax value for many of the newer 45nm Core 2 CPUs. When this value is wrong, HWMonitor will not correctly report your core temperature. Here is what the Intel specs say for a P8400.
IntelŪ Core?2 Duo Processor P8400 (3M Cache, 2.26 GHz, 1066 MHz FSB) with SPEC Code(s) SL3BU, SLB3Q, SLB3R, SLB4M, SLB52, SLG8Z, SLGCC, SLGCF, SLGCL, SLGCQ, SLGE2, SLGFC
TJunction 105°C
To correct HWMonitor, you will need to go into the hwmonitorw.ini configuration file and make sure the correct value is being used. This is a text file so you can edit it with Notepad/Wordpad.
CPU_0_TJMAX=105.0
The ThrottleStop log file shows a problem. When a P8400 is lightly loaded, it will be able to use some Intel Dynamic Acceleration. The default multiplier is 8.50 but when one core is in the C3/C4 sleep state, the other core is allowed to briefly use the 9.00 multiplier. The FID / multiplier being used will rapidly cycle back and forth between 8.50 and 9.00. ThrottleStop accurately reports the average multiplier being used when this is happening. That's why your log file shows FID / multiplier numbers bouncing around between 8.50 and 9.00 at the start of your test when your CPU is lightly loaded.
During your Prime95 stress test, both cores are fully loaded. This leaves you with no Intel Dynamic Acceleration so your CPU will be running both cores at the default multiplier which is 8.50. Your log file shows that even when your CPU got quite hot, it continued to run both cores at full speed with absolutely no throttling so everything was great but.......
After you reduced the load, then a couple of problems showed up. First the multiplier dropped down to 6.00 which externally slows your CPU down to 1596 MHz. It wasn't running at this low speed before you started the test. After you stopped running the test, your CPU core temperature was returning back to normal but the bios decided now was a good time to start throttling your CPU. That's a bug.
After that it also started using clock modulation throttling which is a sly way to slow down a CPU internally without notifying a user why their laptop feels so sluggish, especially when combined with the reduced multiplier. There is no reason for either throttling method to be used on your CPU let alone both at the same time. It's just a sign of a poorly designed and buggy bios.
If you were only using Prime95 during your test then you weren't even working the GPU. This throttling problem can get much worse if you try to use the CPU and GPU simultaneously such as when trying to play a modern game.
Dell's previous generation of laptop, the Latitude D830, had no problem running Prime95 + Furmark to stress the GPU at the same time without overheating or throttling. The newer
Latitude E6400 with the Nvidia Quadro NVS 160M can't run that same test without severe performance throttling which may last well after the load has stopped leaving your computer feeling sluggish and unusable. Randall Cotton's 59 page Throttlegate paper clearly proved that.
For comparison, here's a half hour CPU + GPU full load test on the previous Latitude D830.
Imageshack - furprimetest.png
You can contact Dell but they will most likely give you the endless run around. The reps you talk to are not allowed to admit that there is a design flaw in many Dell laptop models built during the last two years. They don't have a bios available to fix this problem and they are too busy fixing or covering up many newer models with similar throttling issues. If you can get Dell to come out and clean your heatsink and redo your thermal paste, that might help the problem a tiny bit but that won't fix the problem in these machines that is at the bios level. Your laptop started severe throttling when the CPU temperature was fine. The problem runs much deeper than that.