Hong:
As said before, there is no limit for Cp/Cpk values. Imagine a turning process that has a spread (6 sigmas range) of 0.05. Give me a tolerance of 0.05 and you’ll get Cp=1, give me a tolerance of 50 and you’ll get Cp=1000, with the same process.
However, there are some strange things in what you tell. For Cp/Cpk, sigma is calculated from “within subgroup” variation, and that means that you need several subgroups of several parts each (for example, 20 subgroups of 5 parts). Also, stability is a prerequisite for Cp/Cpk, which are meaningless numbers if the process is not stable. That means that you must plot those several subgroups in a control chart and find no out-of-control signal, and that those subgroups should be taken in a time range large enough to let all the possible sources of variation to act on the process. I wonder, how did you do all that with only 8 parts?
By the way, I do not know your process, but a sigma of 0.001 in a characteristic that is 456 is only a 0.0004% of variation over the nominal, which looks too small for a plastic injection.
.
