[alwayssts]

Let's stop. Reflect.

In a benchmark stressing purely tessellation and another benchmark testing mostly tessellation, a gpu that clocks itself dynamically according to total gpu load to fill a 225w tdp performs better in a given unbalanced scenario than one that does not use 225w or do the same dynamic clocking.

Seriously, this boost thing is awesome tech which no one can deny...but it kinda makes certain benchmarks completely bs. Remember PhysX in 3Dmark? Different but similar...your gpu also has to do graphics and is not just using all flops on physx. This is one of those times where it has to be said people play games, not benchmarks.

Let me be clear...nvidia is seemingly ahead in tessellation, but not nearly at the level those benchmarks imply. Also, I question the realistic practicality of that level of tessellation when used in congress with a realistic gaming scenario at any given resolution. I would rather my tdp (if that is the new performance metric/bottleneck) be used on higher flops or pixel/texture fillrates, wouldn't you?

edit: nvidia basically said the exact same thing to legit reviews apparently. Thanks for posting that quote 54thvoid!