FEAATURE//ANALYSSIS
With anti-aliasing, in all its various forms,
we’re looking to find a way of cheating the
lack of resolution by allowing those border
pixels to be not just one or other of those
two final colours but a blend. This softens
the appearance of those edges, making for a
more natural overall look to the image. This
same principle also allows us to reduce the
other undesirable effects described earlier.How does anti-aliasing work?
In computer graphics, there are myriad
types of anti-aliasing algorithms, and we’ll
be exploring the most relevant ones for 3D
graphics later, but the fundamental principle
of all these techniques is the same. Either
they sample the 3D scene (or some portion
of the scene) at a higher resolution and
average out the result to create the final,
lower-resolution image, or they simplysample the final 2D image and average out
the adjacent pixels.
The former technique works during
the actual rendering process, where the
computer knows about the 3D construct of
the scene and can use that information to
its advantage to create a better effect. The
latter, though, happens only after all the
other rendering has been performed, so
it’s just working on a flat 2D image, where it
doesn’t inherently know about the geometry
of the underlying scene. These sorts of AA
techniques are known as post-processing AA.
We’ll discuss each AA type in more detail later,
but first we wanted to set the scene in terms
of how we’ve tested each method.How we test
One of the difficulties when testing AA is the
sheer number of options available and the
fact that there’s significant overlap (or indeed
gaps) in the systems built into a game, what
it offers as options and what can be forced
via a graphics card driver. For instance,
several games offer few conventional
mid-rendering AA methods, as they use
several post-processing effects or deferred
rendering, which means these techniques
simply don’t work.
AMD and Nvidia also have slightly different
terminology for features, as well as some
unique technologies of their own. For
instance, Nvidia doesn’t offer conventional
super-sampling AA but instead has Dynamic
Super Resolution (DSR), while Nvidia’s latest
DLSS technique is unique to the green team.
As such, finding a game that worked
across a wide range of AA techniques, and
offered a meaningful comparison was tricky,
and in the end we settled on an old classic:
Counter-Strike: Global Offensive. This game
supports all the conventional AA techniques,
as well as several of the latest post-
processing techniques.
What’s more, this game also makes it easy
to make comparisons using a low resolution
and a modest AMD Radeon RX 560 GPU. The
low resolution simply makes it easier to pick
out differences (and is still reflective of many
people’s gaming experience), and AMD’s
driver also supports super-sampling AA.
Crucially, the modest RX 560 GPU isn’t CPU
limited at the upper end of performance, so it
shows a very clear change in performance as
different AA techniques are applied.AA enabled
AA disabled
Without AA, edges and thin lines looked jagged and
broken up. With AA, the image is smoothed out
THIS ALIASING CAN
RESULT IN SEVERAL
UNDESIRABLE VISUAL
EFFECTS, THE MOST
OBVIOUS ONE BEING THE
DREADED JAGGED EDGES