While it is true that a PC2700 chip is exactly the same as a PC3200 chip, the statement that PC3200 is overclocked PC2700 is not entirely accurate.
First, let me qualify the following statements by saying that I've worked in the Test department at Micron in Boise.
PC2100, PC2700, PC3200, PC3500. The chips are the same. The only difference is that the higher speed ones have been tested to work correctly at that speed.
Once a wafer has been fabricated, it goes through a process called "Probe". Here, all the individual dies on the wafer are given their first electrical test. They are then marked as either good, bad, or repairable. I won't mention how exactly the parts are repaired, as it might violate confidentiality.
The next step in the process is where all the individual dies are cut from the wafer with a diamond saw. The wafers then go to wire bond where gold wires are bonded from the pin substrate to the die itself. Then, the chips go to encapsulation where they recieve their hard candy-coated shell. Next, the chips go to burn-in. Here they recieve their second electrical test to ensure that they are all indeed working.
After passing burn-in, they then go to machines which do full functional checks. All the parts are tested at various speeds, and sorted accordingly. The parts that work correctly at the highest speeds are sorted into lots with other chips that also work at the highest speeds. Chips that do not work at the highest speed, but do work at lower speeds, are sorted into lots of parts that also work at lower speeds, but not at higher speeds. Thus, you can have PC2100, PC2700, and PC3200 all coming from the same wafer. The difference is that some work at high speeds, some do not.
Why don't all the parts work at the highest speed? There are any number of things that can go wrong during the fabrication process. Imperfections in the lenses used, in the various etch processes, etc.
The reason that we see speed increases in parts is because as time passes, the manufacturer gets better at making the parts. Every single step is recorded, and every single error is tracked. The longer a part is in production, the more time there is to track down problems in chip fabrication. As such, when a particular part starts production, the failure rate per wafer may be more than 50 percent of the total parts on the wafer. As time goes on, and the fabrication process gets better, the failure rate drops. This also explains why prices decrease once something is in production longer. More money is made when 90% of the parts per wafer are good, than when just 50% of the parts per wafer are good.
Interestingly, even parts that do not pass all functional tests can stil be used. These parts go to various manufacturers who don't need certain functions. For example, if a part won't correctly do some sort of error check, it might go to some toy manufacturer whose application does not use the function which does not work.
So, to sum it all up, yes, if you got a PC2700 and a PC3200 stick of RAM, you could very well have the same parts on each stick. The only difference is that one part will be marked with a number such as -8, while the other one is marked with -6 (which is slower than -8)
Some might argue that they have tried PC133 in systems running at PC66, and it didn't work. This is true in some cases. But the reason is not the speed. The reason is that you're probably using a stick of ram with parts of a density too great to be supported by the machine they are in. (See the memory FAQ)
In conclusion, the statement that PC3200 is simply overclocked PC2700 is both true and false. It's true in that the chips on a PC3200 stick may be the same as teh chips on the PC2700. It's false in that the chips on the PC3200 stick have been electrically tested to work correctly at the higher speed.