They are just speed binned. Meaning Intel (or whoever the manufacturer is), tested the processor at various speeds. If it handled 200 MHz, it became a 200 MHz chip. If it failed the 200 MHz test, but passed the 166 MHz test, it became a 166 MHz chip.

Sometimes though yields were good and people wanted to buy more 166 MHz chips than 200 MHz chips and then a 200 MHz chip would become a 166 MHz chip and overclockers loved that 😊

or they were tested and passed using a higher bus multiplier.

Intel or other manufacturers don't put different dies in a single silicon wafer. Dies are first checked to work and then packaged. After this they probably do the "burn in" tests (not sure if they call it like this) and speed tests. So, as Phil says, they test the CPU's at various speeds and go slower if the chip fails the test.

These days manufacturing processes mature so fast that most chips work perfectly at maximum intended speed for that specific micro-architecture.

We may think that this seems very wasteful, but designing a different die for every single variation is simply not viable economically.

I wonder at what point they put the multiplier locks in (for locked cpus). Obviously it has to be after the speed tests, but do they decide right away "this is going to be 166Mhz" or do they store it and watch demand, then at the last minute before shipping out to sell, lock it at 166Mhz when they decide they have too much 200Mhz inventory compared to demand.

Aha, this explains things. Thanks! You learn something new every day... 😎