Alpha screenshots with FXAA and annotation feedback.

However, it's not easy to form an opinion about the effects of different types of anti-aliasing on the image quality in motion with just screenshots. Therefore I will look forward to experiments with SMAA (2Tx) in due time, which in Crysis 3 was received as a good trade-off between image quality and performance. It's specifically the temporal component of SMAA (2Tx) that is lacking in many other anti-aliasing algorithms.

As for average people's perception: Perhaps they do notice the different types of aliasing and the difference between types of anti-aliasing when they see them, but may not be technically articulate enough to describe them accurately. Or they may be too accustomed to visual artifacts in games to complain about them. My point is I think they will appreciate image quality even when they say it's all about the game play.

SMAA in it's various forms are also on my list, though those are going to take a bit more work from Pinbender to implement. SMAA 1x can be injected like FXAA, but SMAA S2x and T2x require engine support which I hope I can convince him to add.

Very Sad that I´m not in the alpha but if I would throw some 24xEdge detect on PA.exe (And other graphic tweaks which are available in the nvidia control panel)

Won't actually work. We have a deferred shading renderer so the regular MSAA techniques don't work.

Thats bad But I´m not in the alpha anyway

Resolution: what can be resolved by the image. This may be pixel counts, but not necessarily. Do not let neologisms replace the meanings of words.

This is why the "average person" absolutely hates FXAA for being "blurry" right? It's because they think it looks good?

Anyone with eyes in their head can see aliasing with only FXAA.

"Releasing a game in 2010 without AA is a completely foreign concept to me. If the technique you're using makes it impossible to use AA then you're using the wrong technique."

Deferred rendering is cancer.

Fair enough.

Deferred rendering is the best option for the job right now. There are a few more techniques for rendering huge numbers of lights, some of which are actually faster than deferred and work with all AA techniques. The problem is some only work on top end hardware from the last 2 years or are too expensive to use on hardware older than that. Basically it means if you want a scene with a few hundred lights you have to have a deferred rendering pipeline at least for the older hardware, and it doesn't usually make sense to support multiple rendering pipelines at the same time if you can help it.

If every gamer was running high end hardware from the last 2 years we could as an industry switch to forward rendering with linked list fragment pages or light indexed tiled techniques. That's not the reality though.

edit: Xbone and PS4 might be fast enough that those techniques might become more commonplace, but I suspect deferred lighting techniques (like Crysis 3, Battlefield 3 & 4, Killzone 3 & 4, etc., which can partially support MSAA) will be the defacto for some time. Because of the extreme number of polygons we're trying to push deferred shading may not be an option.

Please mind your fingers.

I agree. For example Forward+ rendering seems promising. At least, when you're running a new top-of-the-line AMD card. Besides the recent hardware here's another problem, we'll need a vendor neutral solution.

"As many of you know, DiRT Showdown performs significantly better on AMD graphics hardware. This is largely because of the adoption of Forward + and Global Illumination techniques, both of which rely heavily on the DirectCompute capabilities of the GPU and harness the raw compute power offered by our GCN Architecture"

http://forums.techgage.com/showthread.php?t=12329



I wonder if it's A) possible and B) worth it to be able to switch between Deferred Lighting and Forward+/Global Illumination?

If we had a team of graphics engineers (not just pinbender), an mature engine with many years of work in to it and a many million dollar budget like the Dirt teams have, sure.

Only a couple more Kickstarters to go then ^_-