Indeed, it was really fun! I played 2-3 "rounds" of 4 player vs with some friends and then randomly some single player just to see a few of the environments. It was a ton of fun and we all walked away commenting that we want it at work/home nowNice to hear you liked it
Totally understood and agreed that FXAA is (far) better than nothing. Just wish we didn't have to deal with such old hardware these days...We really wanted to achieve solid 60 fps, so MSAA was out of the question from the beginning. For a long time we didn't have any antialiasing solution, but when I got FXAA to run at 0.8 ms (with some branching tricks) it was added. It is surely better than no antialiasing at all, but definitely not something that I would be happy in using in next gen titles (when anything better is possible).
Yeah temporal super-sampling/reprojection is interesting especially for the far geometry as you mention. Again it doesn't really help the 'worst case' (lots of fine occlusion and movement), but it's a somewhat rarer worst case than for other algorithms.Temporal techniques work very well with 60 fps rendering (smaller time step = frames are more similar to each other = more surface pixels that can be reprojected without problems). I did some experimentation with temporal techniques during the project (and during Trials HD project), but I didn't have time to get any production quality results yet.
Yeah it wasn't a criticism, just a note There's only so much you can do in a fixed time/memory budget and I think you've hit a pretty good spot for consoles. It was only rare cases that I noticed it and mostly because I look for it Overall the shadows looked nice and I still can't believe you made EVSM work in 16-bit on a console - tons of props for that, and it really does show nicely when you see a high-frequency shadow projected on the ground. In some cases there's less aliasing in the shadows than on the original geometry, which I find kind of awesomeYes... our SDSM-esque technique doesn't help that much with the worst case performance. ... The limited 16 bit range in the shadow map channels does not help either. With that low precision, the variance calculation results in some occasional (but visible) graining. And the graining differs from cascade to cascade, so it's easier to spot the cascade seams.
Anyways really great work - can't wait to get my hands on it!