Originally Posted by KraxKill
Were your scores actually higher though? I doubt it. Check each Bios clock for clock.
Also I suspect as with the numerous posts above, higher clocks don't always to translate to better performance.
My card is on water and shunted, and though I can run the card at 2100 in Firestrike/Heaven, 2138 in Valley, my best performance is actually at 2075.
My situation could be due to me running a 4790k but it's at 5.0ghz so it isnt a slouch.
Also, don't forget to all those that are benchmarking. Not all tests/loads are power limited. The higher the resolution/complexity, the more likely it will be that the card is TDP limited.
In addition, the extra TDP headroom doesn't help your average FPS as much as you might think. What it does help a lot is the minimum
FPS. Even in heaven, there's only like 5% of the benchmark that's really intensive, the rest of it won't power limit even with a 300w cap anyway.
Try running The Witcher 3 in 5k res with everything maxed, and watch your minimum FPS through a benchmark run.
Switching P-states does have a slight latency to it and can cause a bit more irregularity in frame times if they are bouncing around due to coming up against the TDP limiter.
Being able to stay in the absolute top P-state and max clocks throughout an entire benchmark run improves the perceived smoothness from my experience as well, even if it might only translate to a .1 difference in the average FPS at the end of the run.