Fun topic. Binning (selling by different SKUs) for a given IC is very common. Has been standard in the industry for 30+ years. This is partially because reticle sets (also known as a photomask set) are VERY expensive. So - they can sell the same product at different price points. As noted, this also allows them to sell partially defective products as 'good'. This is most common on 'complex' (high transistor count) chips. Most common is memory capacity, processor speed, and available core count. They test (validate) all parts of the chip. As a general rule, only testing part of the chip isn't useful. Now where it gets interesting: Typically early in a process cycle (when stepping into a new minimum feature size), yields are poor and so binning is 'legit'. E.g. If they mark it as an X GHz product, then you can probably get away with X*1.2 Gz if you were to unlock it and make sure you have better than mfg recommend cooling (for example). If a chunk of memory is locked out, then it is very likely locked out for a good reason. BUT ... as a the process cycle is matured, then yields generally go up. And, in some cases that will result in false binning. If customers are still demanding an X Ghz product at a given price point, then they will take better bins and classify them at the lower speed / capacity / etc. Obviously, these yields are VERY closely guarded secrets and unless you are an industry insider you will not know where a given company is on a given process cycle maturity. But, yeah... If this isn't a mission critical system ... go unlock and see if it crashes.