|Duh Voodoo Man's PowerLeap iP3/T Overclocking Experiment|
|SQUEEZING THE TUALATIN CELERON UNTIL IT HERTZ...|
CRANKIN' IT UP:
Click on image to enlarge WCPUID @ 117MHz FSB
I began by installing and running benchmarks on the original "rev.1" iP3/T and Celeron 1.2GHz at 100MHz FSB to establish a baseline, followed by the rev.2 rig with the 1.3GHz Cely. As expected, the 1.3GHz processor was slightly faster than the 1.2, with benchmark result increases ranging from close to nil up as high as 8.5%, depending on how heavily CPU dependent the given test was. I would expect an overall "real world" performance difference between the two Celerons to be somewhere in the 3-5% range. Your mileage may vary, as they say....
Now it was time to get into the really interesting stuff. I removed the rev.2 iP3/T-Celeron 1.3Hz rig and jumpered the core voltage up to the 1.60-1.625v setting, reinstalled it in the PC, and then boosted the FSB speed in the BIOS to 117MHz. I was tempted to begin at the next lowest FSB setting of 112MHz, but opted to start more aggressively. The gamble paid off, as the unit booted right up into Windows and appeared to be running normally. A quick check with WCPUID confirmed the FSB speed setting (within measurement error limits) and showed the processor clock to be at the expected 1.52GHz level (click on image at right to see the numbers).
I proceeded to run the set of benchmarks with no problems. Though the system only ran at 1.52GHz for about a half an hour, it appeared to be perfectly stable. Benchmark results are summarized on the next page. So far, so good!
Encouraged by the success of this first overclocking attempt, I rebooted the system and raised the FSB setting in the BIOS to the next available value--124MHz. If successful, this bus setting would translate to a CPU clockspeed of 1.61GHz, a very respectable overclock. The system POST'ed uneventfully and the first part of the Windows load ran normally, but the machine froze when it got as far as the initial appearance of the Windows desktop. Suspecting that it needed a bit more "juice", I powered down, pulled the processor rig, and jumpered the core voltage up to 1.70-1.725v. Then, I reinstalled it in the PC, pushed the power button and held my breath....
Click on image to enlarge WCPUID @ 124MHz FSB
SUCCESS!! The machine booted all the way into Windows and appeared to be running normally. Again, a quick run of WCPUID showed the expected bus and processor clockspeeds (image at right).
The benchmarks again ran well, with the exception of a single kickout to the desktop with the PCMark 2002 test. This appeared to be anomaly, though, because I repeated the test two times successfully, and none of the other tests--including the demanding video benchmarks--showed any problems. Again, results are shown in the tables on the following page.
With the immortal words of Inspector "Dirty" Harry Callahan in mind--"Do you feel lucky, punk?"--I rebooted and entered the BIOS once more, raising the FSB speed to the next available setting of 129MHz. This would produce a processor clockspeed of 1.68MHz if successful, but I had real doubts of the ability of the Tualatin Celeron to reach that level, based upon what I'd read.
It turned out that those doubts were well founded. On first bootup, the system made it through POST but froze very early in the Windows load process. I powered down, pulled the iP3/T out, and jumpered the core voltage up to the maximum 1.80-1.825 setting. This time, the PC booted all the way into Windows, but was obviously unstable. I got a couple of odd failure messages as Windows completed loading, and none of the benchmarks except the two Sandra tests would run all the way through without kicking back to the desktop or causing the machine to spontaneously reboot. The 1.68Ghz level was evidently more than this CPU could handle, at least without going to special cooling measures that I was unwilling entertain.
So, with this particular CPU and motherboard combo, the end of the line turned out to be 1.61GHz. I suspect that I could have squeezed a few more MHz out of it if I'd had an FSB setting option between 124 and 129MHz, but such was the constraint with the AOpen AX6BC. Still, a 24% overclock ain't too shabby, and confirms that the 1.3GHz Tualatin Celeron is a decent overclocker.
So let's take a look at the actual benchmarks results....