Custom PC's Ultimate Core 2 Overclocking Guide

Posted at: Wednesday 31st October 2007 by Alex Watson

Not sure where to begin when overclocking? Alex Watson uncovers the dark art of coaxing more speed out of your PC's components for free.

The chips are then sorted according to the models the company is selling. As a CPU's speed is a somewhat artificial number, there will clearly be a little room for tweaking. The second factor, which creates a bigger overclocking opportunity, is Intel and AMD's need to make CPUs in large volumes in order to offset the huge cost of their fabs. Intel's Fab 24-2 in Ireland, which makes 65nm CPUs, cost $2 billion to get up and running, for instance. Manufacturing CPUs in volume means that over time, the fabs perfect the process of producing certain types of CPU core, so the yield of chips with higher clock frequencies increases.

In theory, this means that the price of a top-end CPU should drop considerably, but Intel and the firms that build PCs around premium chips don't want this to happen. They want to preserve the premium feel of high-end products, and they don't want people to hold off buying a product because they know the price will plummet. Keeping high-end chip costs high means that there's still demand for slower, cheaper chips.

By this point, however, the fab is expert at making the fastest cores required, so the only way to meet this demand is to classify some faster-CPU cores as slower models. This is the case with the Q6600 at the moment; the box says 2.4GHz, but it's more than happy to run far faster, as it came from batches of CPUs capable of higher speeds.

The main problem is determining which chips will be good overclockers, as there's no obvious way to determine this - Intel doesn't write '2.4 GHz (but actually it will do 4GHz easy-peasy)' on the box. Luckily, it's possible to get an indication of which CPUs might overclock well by considering certain core revisions and steppings.

Core blimey

During the life cycle of a CPU family, Intel and AMD make changes to the architecture in order to fix bugs, tweak performance and add new features. These revisions, which are identified by a unique code or number, are called steppings.

Minor revisions to a core's architecture are generally identified by a different number, while major revisions are usually indicated by a letter. These changes are often very beneficial to overclocking, and the 'G0' Q6600 is a great example of this. The original Q6600 had a stepping code of B3. The move to G0 has seen the TDP drop from 105W to 95W, so the chip runs cooler, and its 'T-case' temperature has been raised by 9˚C, indicating that the chip can sustain a higher temperature (and therefore voltage and/or overclock) than the B3-stepping chip before becoming unstable. These changes clearly indicate that the core is more thermally robust; given that heat is a key challenge for an overclocker to overcome, this is extremely good news.

Clearly, if you're interested in overclocking, tracking down a G0 chip is a good start, and retailers such as Scan and YoYoTech are now differentiating between stepping codes on their websites. You can easily see which stepping your CPU has by using CPU-Z (www.cpuid.com/cpuz.php), although there's something far more useful you can use if you haven't yet bought a chip: the sSpec number on the box.