Overclocking the 500

366 all over again?

In our Celeron 366 to 550 overclocking article, we explained how the most recent batches of Celerons are manufactured to tighter tolerances in order to push out more chips capable of running at 466MHz, 500MHz, and above. Most likely, the Celeron 500 is the exact same chip as the new 366, especially those out of the same Malay fab. Assuming this, we eagerly took to overclocking tests after running through our regular regimen of benchmarks.

What exactly did we try?

After smearing the top of the CPU with a (tiny) bit of thermally conductive paste and attaching a hefty heatsink, we threw the lone Celeron back into our Abit BP6 dual-Celeron motherboard. While we weren't testing dual processing, the BP6 is an excellent overclocking choice as it allows for 1 MHz and 2 MHz increments at critical frequencies (around 75MHz and 83MHz), which allowed us to test the absolute limits of the processor's overclockability.

First, we booted up to confirm that 500MHz worked as advertised. Then, moving to the BIOS SoftMenu II, we set the bus speed to 75MHz, one standard notch above 66MHz. This resulted in 563MHz, which POSTed beautifully and burned in for several hours without fail. The next step was 625MHz at an 83MHz FSB, which tends to be a dangerous frequency due to the strain on the PCI and AGP bus. Unfortunately, the system crashed repeatedly, even at 2.3 Volts. All of our testing components have been confirmed to function properly at a 90MHz bus setting, so apparently our chip failed the 83MHz test, completely ruling out the possibility of 750MHz (7.5 x 100), which we tried anyway.

All is not lost!

This is where the fine adjustment capability of Abit's latest SoftMenu 2 shines. We then worked back from 83MHz in 1 and 2MHz increments in order to determine the absolute top speed at which the processor could remain stable. At 82MHz (615MHz), the Celeron would POST and boot up, with random crashes during Win98 loading or benchmarking. 81MHz was not available, so we moved down to 80MHz, an even 600MHz. Here, the CPU stayed rock solid, even after several hours of testing.

600MHz is generally accepted as nearing the manufacturing limits of the Deschutes/Mendocino core, and the .25um process as well. The heatsink was never more than 24oC, while the BIOS thermistor reported a crisp and chilly 32o C directly under the processor core. Since neither heat nor peripherals are evidently issues in our overclocking tests, it's safe to assume that at the current temperature, our CPU has reached the end of its clock speed line. Now, if we could supercool the processor down to -40o C, it's theoretically possible to gain an additional 20% increase or so, as CMOS components run faster when cooler. This however, isn't a realistic solution for most users.

Small print

Obligatory reminder time: our overclocking tests were done with only one CPU. This means that out of all the thousands of other CPUs available, it's not a large enough sampling to claim that x% will be able to reach 563MHz or 600MHz. When the Celeron 500 is released, we'll have more information on what overclocking yields will be, but for now, take this report as an individual test on an individual chip - your results may vary.