Yes there is a performance hit associated with it. Tho in many cases that performance hit would be less than just using much more detailed base models.

You ultimately have to draw the polygons regardless.  The way tesselation is used in current tech demos and games just repurpose polygons.  You don't need 100k polygons on a model while you play a game, but you would during a cutscene.  Conversely you wouldn't need that many polygons in the background during these cutscenes and you can draw them into that 100k model and dumb the model down to 50k while playing and drawing more polygons in the background.  Depends how you want to subdivide the polygons and set control points for them.

These should help explain what tessellation can be used for

Hummmmm... really? Same anwser... GPGPU will not be used (at least not in this new generation) and particularly there is no change of water for wine because of GPGPU... there is no miracle... no game running or looking better because the GPGPU.

The others points in the article was expected.

lol you and me agree.

So why you said my name before.. lol lol... I never defended GPGPU.

PS. I'm a little drunk now.

wow that tesselation stuff is intense

Sounds awesome cant waite for the ps4.

It will be used, UE4 uses DirectCompute aka compute shaders (Direct X's GPGPU API) to calculate it's voxel based lighting volumes for example. But mostly it will be used for a few gimmicky effects like fluid/cloth simulation or particle physics. I am sure Nvidia will port PhysX to the GPU on next gen consoles in some copacity, mostly to encourage devs to implement it in cross platform games.

But definitely not the offloading CPU work to make up for weak CPU nonsense that some people seem to think. And really it doesn't apply in the same way in the console space as it does in the PC space, devs already have low lovel access to the GPU hardware on consoles so GPGPU doesn't make much difference in that space exept for slightly more capable hardware.

I feel that Nintendo's perspective on GPGPU is doing things like vertex processing on the GPU instead of relying on the CPU to help out. In comparison to what is presently done with the current console generation, doing things like animation etc on the GPU could be called GPGPU in comparison to current console paradigms so I guess they aren't exactly lying even if the truth is less spectacular.

We should also remember that a lot of what the GPU currently does was once the domain of the CPU and with respect to this change in processing model we do have GPGPU by old standards because as the GPU has picked up more workloads we've grown accustomed to it.

as someone who has done CUDA programming, I can tell you that GPGPU programming no easier than using cell. Objects needs to be copied to the GPU's memory, which could range from fast but small local memory to the slow but larger shared memory. When you copy back, the memory needs to coalesced to achieve optimal write performance. Not to mention, you need to manage and sync warps (cluster of threads) which can cause huge bottlenecks if you require all data before sending stuff back to the main memory.

Of course, this is all from 3 years ago.. not sure how GPGPU programming changed these days.