Questions about Anti-aliasing

Hi, I'm attaching two renderings of the same location, (A) using anti-aliasing (3×3, depth 2, threshold 0.1), and (B) by simply taking a high resolution (just under 4×4 the target resolution) and downsampling the result. Variant B not only finished in less than a third the time, but also resulted in a picture with far fewer bright solitary pixels, which were the reason why I saw a strong need to anti-alias this image in the first place.

Also, I like that variant B gets 'darker into the distance', whereas variant 1 seems to somehow betray the notion that the gradient should recede into infinity. But that's a matter of personal preference, so I'm wondering: Would anybody actually prefer variant A? For those who don't, is there any other reason to use anti-aliasing?

Perhaps I just got a bad example, but I remember to previously have experimented with escalating anti-aliasing settings without being overly successful, and not even really understanding if 2×2 at depth 3 was simply 8×8...

(A) Anti-aliased:

(B) Resampled:

I used anti-aliasing for a while but never do now. I render in a similar way to your method B. Each side 3 times larger than the target size, then shrink it back down in Photoshop. Then add a little sharpening if necessary (usually not).

Not sure if this is faster or slower than anti-aliasing as I haven't tested it. I just prefer the result. Maybe your 4x4 would produce an even better result, I will try it.

I'm assuming you're talking about the anti-aliasing feature used by render to disk? This should perform better than just downsizing a larger image in Photoshop: it performs more advanced downsampling and also subsamples pixels only when needed, giving a speed advantage. However, I won't disagree with you that the second image looks better.

Can you send the UPR to me via email so I can try to find out what's going on?

Yes, I was referring to the render-to-disk anti-aliasing. The parameter set is the one I posted here for the CPU benchmarking.