Prior to introducing our final test results, I've included some screenshots exemplifying my test methods which include running CPU-burn and Prime95. I looped the latter to stress the CPU creating LOAD conditions, from which temps were then recorded. The SF-609 uses a front LED readout; all other utilities are displayed in Windows. The screenshots below show two temperature extremes, but not the most dramatic extremes which would be Idle at 200FSB (1.33Vcore/default), and LOAD at 267FSB, (1.60Vcore). I wanted to show the CPU temp at 4.0GHz while pushing 1.60Vcore.
This overclock wasn't completely stable, and while I may have been able to stabilize the system by raising the Vcore above 1.60V, I didn't think it was worth the potential for damage. I'm not sure if it's possible to "overclock conservatively," I'm at the point where I weigh the potential loss of hardware simply to attain a few more MHz. After all overclocking always seems to come down to wanting, "just a few more MHz.
Unfortunately the P4C800E-Deluxe (or at least Asus Probe) doesn't monitor its PWM (CPU power circuitry) and the S478 board was not originally designed to support a .09-micron core processor. While the Prescott runs at a much lower Vcore (1.33V), its current requirements are much more demanding then its .13-micron older sibling Northwood. The end result has been severe overheating on the Mosfetts and circuits supplying power to S478 CPU's. In my article Prescott effects on motherboard power circuitry
PWM (processor power circuits) temperatures, as measured on the Abit AI7 ran as high as 72°C/161°F and even higher.
Below is a Sandra memory benchmark as I wanted to show, that while running at 4.0GHz the NexXxos XP is maintaining some excellent temps.
Comparatively, this Sandra score is excellent given the fact the system is running at 213MHz off the divider, and with PAT disabled. At 4.0GHz the VAGP must be raised to 1.7V, on the P4C800E-Deluxe otherwise the system isn't stable. Whether this is a result of the divider's failure above a certain FSB speed, I've yet to determine. Final test results:
The graph below shows the highest temperatures which were reached when stressing the CPU 100%, ambient temp was recorded with the SF609.
The rise over ambient temperature is really minimal, even at 4Ghz (a speed not obtainable with air cooling) its only a mere 15°C, impressive!