The power thing doesn't quite work the way you're saying
My understanding of this is as follows:
Basically, look at this chart again:
What you do is there is a MAX TDP (power draw) for a card limit. You can adjust the slider up/down to set the maximum TDP draw.
At stock, the 6950 has a maximum of 200W draw apparently. In the Perlin Noise synthetic benchmark, the core clock adjusts up and down from 650-800 (average around 700) in order to keep the power draw under the maximum of 200W.
Now, when the power draw is increased 5% to 210W, the core clocks can reach higher MHz because the card is allowed to draw 10 more watts of power, hence the average core clock now shoots up to 750-775MHz. Likewise, the FPS result jumps from 140 to 155, a pretty significant jump in a synthetic benchmark.
Once the card is incresaed 10% to 220W, the core clock maintains itself at 800Mhz the whole bench.
To 'overclock' the card, you still have to set the MHz in CCC - so you can tune it to 850W, and that's the maximum clock the core will reach.
Basically it's like turbo boost for CPU's, but somewhat in reverse.
Basically:
TDP Slider - Sets maximum power draw of the card
Core Clock - Sets maximum core clock of the GPU
Thus the card dynamically adjusts the core clock up to the max core clock as long as the overall power draw of the card is below the TDP slider.
That's why I don't take much stock in the synthetic benchmarks, since the leaked 6970 perlin noise FPS figure is very low compared to the 5870 - in all likelihood, it's showing similar results to the chart above in that it's not consistently at 880 MHz core across the benchmark. Games might not be affected as much, but interesting applications for this would be to downclock the GPU for games that already run at 200FPS to save power when gaming. Now that would be sweet.
My understanding of this is as follows:
Basically, look at this chart again:
What you do is there is a MAX TDP (power draw) for a card limit. You can adjust the slider up/down to set the maximum TDP draw.
At stock, the 6950 has a maximum of 200W draw apparently. In the Perlin Noise synthetic benchmark, the core clock adjusts up and down from 650-800 (average around 700) in order to keep the power draw under the maximum of 200W.
Now, when the power draw is increased 5% to 210W, the core clocks can reach higher MHz because the card is allowed to draw 10 more watts of power, hence the average core clock now shoots up to 750-775MHz. Likewise, the FPS result jumps from 140 to 155, a pretty significant jump in a synthetic benchmark.
Once the card is incresaed 10% to 220W, the core clock maintains itself at 800Mhz the whole bench.
To 'overclock' the card, you still have to set the MHz in CCC - so you can tune it to 850W, and that's the maximum clock the core will reach.
Basically it's like turbo boost for CPU's, but somewhat in reverse.
Basically:
TDP Slider - Sets maximum power draw of the card
Core Clock - Sets maximum core clock of the GPU
Thus the card dynamically adjusts the core clock up to the max core clock as long as the overall power draw of the card is below the TDP slider.
That's why I don't take much stock in the synthetic benchmarks, since the leaked 6970 perlin noise FPS figure is very low compared to the 5870 - in all likelihood, it's showing similar results to the chart above in that it's not consistently at 880 MHz core across the benchmark. Games might not be affected as much, but interesting applications for this would be to downclock the GPU for games that already run at 200FPS to save power when gaming. Now that would be sweet.
.
