Will fillrate take a backseat as shader based AA gains popularity?

[jhokie]

As shader based AA methods continue to gain steam, will we see an end for the need of high fill rates in cards (aside from multi-monitor)? Will amount of rops stop increasing and more unified shaders/cuda cores get crammed on cards?

Don't get me wrong, I would like to have MSAA as a viable option so I can compare it to FXAA or MLAA in my games. I do have to say, I really haven't noticed much aliasing in Deux Ex (though the textures and other effects are clearly not up to cutting edge standards). I can only assume that shader based AA methods will continue to evolve and look better and better. Clearly this method does main a minimal impact on performance compared to 4X MSAA or greater.

I would imagine this next series of cards may still see a substantial increase in fill rate, as they will have to continue to support a varied amount of AA options. But as shader based methods continue to spread over the next several years will this be the end of traditional AA as we know it? 2 years from now is there any benefit to be gained from pushing 150 GP fill rate in a crossfire setup if we're not using any MSAA or SSAA? Is the majority of the work going to be done by the shaders?

Perhaps someone with more technical knowledge can chime in.

 

[NaturalViolence]

As shader based AA methods continue to gain steam, will we see an end for the need of high fill rates in cards (aside from multi-monitor)?

Define "fill rate". I assume you mean ROP pixel fillrate.

Will amount of rops stop increasing and more unified shaders/cuda cores get crammed on cards?

Nvidia has already starting moving towards a more compute (GPGPU) focused design with fermi and amd is planning to follow the lead very soon with graphics core next.

Don't get me wrong, I would like to have MSAA as a viable option so I can compare it to FXAA or MLAA in my games.

.....you're assuming MSAA is always fixed function and FXAA/MLAA is always shader based. When in reality a lot of games use shader based MSAA.

I can only assume that shader based AA methods will continue to evolve and look better and better.

Again, you're talking about a move from render time processing methods to post-processing methods. This has nothing to do with shaders vs. fixed functions.

But as shader based methods continue to spread over the next several years will this be the end of traditional AA as we know it?

SSAA is still around despite MSAA being the favored approach by game developers. It can never completely disappear.

2 years from now is there any benefit to be gained from pushing 150 GP fill rate in a crossfire setup if we're not using any MSAA or SSAA? Is the majority of the work going to be done by the shaders?

That's what eyefinity/nvidia surround is for.

Is the majority of the work going to be done by the shaders?

Nearly everything related to graphics in any way is done with shaders these days.

 

[defaultluser]

I think the importance of dedicated MSAA hardware will fall-off, but otherwise nothing will change. ROPs are still required to render the final element to the screen.

I suppose if resolutions remain fixed, then we may not see ROPs increase with new generations of video cards. I guess we'll need to wait and see