It seems to me you are implicitly assuming that power and computation speed have a perfectly linear relationship. I don't see how that is possible. Even if the processor clock rate and power consumption have a linear relationship (and I do not know if that is true), there are other factors involved in computation, such as memory and disk access. Since those don't get faster with the turbo clocking, and since they tend to be bottlenecks relative to the processor in any case, I don't think increasing the clock rate is going to give you a linear increase in overall performance. It will just increase the percent of the average compute cycle that you end up waiting for data that isn't in cache. I may have this a little bit wrong, since my knowledge of computer architecture is extremely dated ("OK, Boomer!"
); but the point is, due to nonlinearities in the relationships between power and compute rate, and possibly even CPU clock rate, that formula above is way too simplistic to support your main point.
