There is redundancy so that if there are failing transistors found during manufacturing testing, the bad circuit block(s) can be removed from operation by blowing fuses. For a microprocessor, that is mostly limited to fusing out whole cores or large sections of cache memory.
If a transistor fails later on, during operation, you are generally screwed.
Your second question is harder to answer. If by “fail” you mean what we would normally consider a truly non-functional transistor, that doesn’t normally correlate with speed. But transistors/circuits don’t have to be “hard” failures; there are other failure modes that can lead to speed degradation.
*Source*: I’m a former microprocessor reliability engineer.
Latest Answers