Size does matter
Physics & Thermodynamics 101
Ok, we can all name circumstances where size does matter, and where it doesn't. This isn't the place to argue over what belongs in each category. What I will say is that size does matter for heatsinks. The physics for this are simple. Objects with larger surface areas can dissipate more heat, and this is exactly what you want your heatsink to do for your CPU. Of course, there are exceptions concerning efficient design and materials use, but all things being equal, bigger is better. The material that a heatsink is made of also determines its ability of absorb heat. Combining these factors together, gives us a measure to compare heatsinks, heat capacity.
Heat capacity is normally measured in degrees Celsius/Joule, the standard comparison is done is Celsius/Watt, where a Watt is a Joule/second. Generally these two numbers are comprable, as the amount of time it takes for heat to diffuse across a good conductor is small. It gives the increase in temperature of the heatsink for every Watt that is absorbed from the environment, in this case, it is your CPU. A major factor in comparing these numbers, requires a comparison on the fans mounted to the heatsink. You can make a cheap heatsink as well as the Alpha by putting the biggest and baddest fans on the cheap heatsink, while taking the fan off the Alpha.
The Alpha P125C has two ratings of heat capacity, one for 60cfm of airflow (0.245 C/W) and one for 40cfm of airflow (0.340 C/W). These numbers are very good, and are not just theoretical numbers as the thermal paste that is provided allows for an efficient energy transfer between the heatsink and CPU. So there is a function to this form!
So does a cooler running CPU guarantee greater overclocking? Yes and no. While lower temperatures do make semi-conductors run faster, other factors such as trace precision and quality have a say in overclocking. For example, no amount of hi-tech cooling can make your 486 run at 300MHz, the trace element widths prohibit it. So, there is a theoretical limit of processor speed, dependent upon trace widths. A trace width of 0.25 microns has a speed limit of about 600MHz, for the Pentium II core. Faster CPU's may exist, but these are rare.
Welcome to the 300A
We then come to the CPU that carried "overclocking" into the social dictionary, the Intel Celeron 300A. This CPU was released a few months after Intel had perfected the 450MHz processor, after Intel had verified that it's manufacturing prowess would support 450MHz systems. As the top Intel processor at the time was an Pentium II or Xeon 450MHz, it made little business sense to market the Celeron at its potential, but rather at its competitor's potential, 300MHz. Although the multiplier has been locked on Intel CPU's for a while, the 4.5 multiplier on the Celeron 300A made the transition to 450MHz, as simple as upping the bus speed to 100MHz. That's what many of us did, even without exotic cooling methods.
The 300A that we tested ran at 450MHz, 2.1Volts perfectly stable, with a generic heatsink and fan combo, with only a thermal pad. This CPU ran painfully hot; touching the backside of the CPU after a series of 3D games would induce pain reflexes that no amount of concentration could counter. I even went so far as to cut up a Pentium heatsink and attach it to the backside. This didn't work to well as the heat from the CPU weakened the thermal tape used to attach the heatsink, sending the heatsink plummeting towards my Abit motherboard. Standard PC100 SDRAM was used for the test.