thats funny, my professor for my computer organization and design class was just talking about that. she was saying that the cpu's are basically made all the same way its just that they are separated by "quality". they higher the "quality" of the chip the more they sell them for. same for the lower quality. and after the ships have been in production for so long, intel for example is able to improve their production of the chips making more of the higher quality ones and then they are able to lower the prices for the consumer. (she said something like that i dont remember exactly, i might have said something wrong)
That's how it works, in theory. But once Intel commits to offer certain products (such as the Celeron D) it must continue to make CPUs for that product segment. Let's say Intel makes 10% of their CPUs with faulty cache (this is quite erroneous, but it is just an example), and that in one day they make 10,000 CPUs. That would mean that out of the 10,000 CPUs made, 1,00 of them would become Celeron Ds. Now that's all fine and dandy, but what happens when the consumer purchases more Celeron Ds than Intel can get out of faulty cache CPUs? Well, one thing is for certain, Intel is not going to deny a potential sale just because they don't have enough rejects to sell as Celeron Ds. What happens then is that Intel will have to start disabling cache on GOOD CPUs in order to keep up with the demand of Celeron D sales. It's all about supply and demand. If Intel can't generate enough faulty cache CPUs, they will turn good cache CPUs into Celeron Ds just the same to keep up with demand. After all, for Intel, a Celeron D costs just as much to make as a Pentium 4, so they may not be making as much profit PER CPU, but they're still making a hell of a killing by selling LOTS of cheaper CPUs. A good example of this can be found on the AMD side. Back when AMD had their Socket A Duron with only 64kb of L2 cache, there were people that were able to "hack" the Duron into unlocking all the L2 cache (256kb) as it's big brother Athlon XP. Not every Duron was stable with 256kb L2 cache, but some were fine, indicating that AMD had deliberately downgraded an Athlon XP-quality chip in order to keep up with the demand of the low cost Duron CPU.
The same thing can be seen in clocks speeds as well, which is something that overclockers count on. A good example would be the Pentium 4. At the same time Intel was selling 3.8GHz Prescott P4s, they were also selling a 2.13Ghz Celeron D. That's quite a bit difference in clock speed considering both are on the EXACT SAME core. It would be very unlikely that the Celeron D 310 was sold with a 2.13Ghz clock speed because it failed to run any faster. The top stable speed of mass produced Prescott CPUs would probably fall into a bell-shaped curve, and if we were to take the Celeron D 310 as being the absolute lowest speed a Prescott CPU would achieve, then it would fall on the far left of the bell curve, meaning that less than 2% of all the CPUs Intel produced would of such poor quality to be used as Celeron D 310s. That's not a whole lot of CPUs to feed the Celeron D 310 product demand, and as a result many CPUs that were capable of attaining much higher frequencies were downrated to Celeron D 310s in order to keep up with consumer demand.
That explains why a lot of Celeron D 310s, 315, 320s and 325/6s can often hit a 50% frequency overclock, and most of the time on stock voltage. Intel had to keep up with the demand of these cheap chips, and so a lot of 3.2-3.6Ghz capable CPUs were sold as 2.13 to 2.53Ghz Celeron Ds. The whole "CPU grading/speed binning" theory is nice and all, but it's not how it always works in real life.
50% OC Cel D - that would be the reason why a 2.4 Cel D I once had was able to be OC'd easily with no Vcore adjustment to 3.6GHz.