The Celeron 566 was the first CPU to undergo testing, and the default Intel heatsink/fan was used for the initial overclocking. Surprisingly, after setting the virtual 100 MHz FSB jumper, the C566 booted immediately to the 850 MHz speed, and at the default 1.5V voltage! This was unprecedented in previous CPU overclocking, and such a large jump in core speed almost certainly requires some sort of core voltage increase as well. The C566 was extremely stable at 850 MHz and with this successful overclock, further FSB increases were definitely required. The subsequent 886 MHz (105 MHz FSB) overclock wasn't a problem, but we were forced to move the core voltage to 1.65V to ensure total system stability. The C566 overclocking progressed all the way up to 931 MHz (110 MHz FSB) without any trouble, but at this point the CPU core heat became a serious issue. The Intel heatsink was getting extremely hot and the default cooling fan was simply not sufficient. Moving to a more robust heatsink cleared up the problem, and by raising the voltage to 1.8V, the C566 was able to maintain 952 MHz (110 MHz FSB). The processor would also boot up at 976 MHz (112 MHz FSB) and 991 MHz (117 MHz FSB) but would crash while performing tasks in Windows. It was even able to boot to 1015 MHz (120 MHz FSB) but then proceeded to lock up during the initial OS load.

The Celeron 600 was a bit of a different story, and initial attempts to hit 900 MHz (100 MHz FSB) were greeted with a totally blank screen. Nothing seemed to work, and I was forced to manually adjust the voltage settings on the Slocket to 1.8V before it started to respond. The PC then booted up at 900 MHz, but was not stable while in Windows. The 810 MHz (90 MHz FSB) and 855 MHz (95 MHz FSB) were easily reached using a lower 1.6V setting, but any higher FSB required a large increase in core voltage and a lot of faith. I then tried a period of “burning in” and that helped stability a bit, but the processor would still heat up to extremely high levels. In the end, I jury-rigged a large heatsink to a high-speed 60x60 fan and that ended up keeping the CPU temperature within tolerable limits. No amount of cajoling or additional voltage tweaking would allow the C600 to reach any higher speed than the 900 MHz overclock.

The Celeron 600 testing was also where a potential issue with the FC-PGA design came to light. Sometimes we take for granted the options present on these Slocket adapters, especially the voltage controls. When trying to reach 900 MHz with the FC-PGA board, there are no hardware voltage controls to speak of, and when changing the hardware FSB to the 100 MHz virtual setting, some motherboards actually try and re-initialize the processor. This can reset the core voltage to default levels (1.5V) and make certain processors extremely difficult to overclock. I've had this happen on other VIA and i810E motherboards, but since the boards were Slot 1, I just hammered down hard with the hardware Slocket jumper settings. No matter what was tried, the FC-PGA board would not allow the Celeron 600 to overclock to 900 MHz, or even boot up at the 100 MHz jumper setting to allow the lower 90-100 MHz FSB speeds to be accessed.

With the Celeron 700, you shouldn't really expect a lot of headway in terms of overclocking. Its 10.5X clock multiplier ensures that even the smallest FSB increase will be felt in much larger increments by the CPU. It was a bit surprising then that the Celeron 700 ran seamlessly at both 788 MHz (75 MHz FSB) and 875 MHz (83 MHz FSB) at core voltages of 1.6V. It also comes within a hair of maintaining a speed of 952 MHz (90 MHz) at 1.8V, but will crash a few minutes after Windows loads. Increasing the voltage or adding additional CPU cooling does nothing to alleviate the problem, and it would seem that 952 MHz is just a tad too high for this core.

While it's difficult to draw firm conclusions with just this small amount of overclocking data, there are a few constants that were easy to see throughout the testing. First, all the tested processors would easily reach 850-875 MHz speeds (if available) at extremely low (1.5V-1.6V) core voltages. The overclocked processors ran perfectly, exhibited no problems at all, and CPU core heat stayed within limits that were manageable using a retail heatsink and fan. Moving higher than 850-875 MHz brought a whole new set of problems and required extreme (1.8-1.85V) increases to core voltage, which resulted in increased core heat (44-46 degrees C), and required upgrades to the CPU cooler.

From this data, the Celeron 566 would seem to be the overclocking chip of choice, given that at the 100 MHz FSB, it hits a very manageable 850 MHz core speed. If you can find it, the Celeron 533A is also an excellent bet, and online polls have it as a virtual lock at 800 MHz. Moving higher than that gets more risky, as our difficulty achieving 900 MHz with the C600 can attest. The 633-700 MHz models have increasingly higher multipliers and aren't especially suited for 100 MHz FSB overclocking.