Earlier quoted context omitted.
The performance will only be better for the first 30 or so seconds anyway before the cpu has to start thermal throttling.
Do you have benchmarks that show anything like this? Or are you referring to the firmware bug that was fixed shortly after release, last year?
The issue last year was because Apple was applying the thermal throttling in such a way as to make the average performance of the CPU significantly lower than it needed to be by throttling too much, too quickly then recovering to too high of a clock speed too quickly thus ping-ponging between a very fast/high/hot state and a very low/cool state. In many cases the CPU would drop down below 600mhz for me for periods of time.
A proper setup will have a graph with a brief (~30s) burst to turbo max followed by a thermal throttle slightly below what the system can sustain followed by a recovery to the max sustained clock speed at just below the thermal limit.
e.g. 4.2ghz for 30s to 2.2ghz recovering to 2.6ghz sustained.
The exact profile is highly workload dependent but a properly designed system invariably follows that general profile.