|
Posted by Thomas
McGuire on December 26, 2001
Manufacturer: Creative
Labs Europe Product: 3D
Blaster GeForce 3 Titanium 200
Check
for the lowest videocard
prices.
The
GeForce 3 was the first Graphics Card to support
Multi-Sampling FSAA, as specified in DirectX 8 & the
GeForce 3 Titanium 200 continues this tradition. The main
benefit to this is that of performance. Modes available are X2,
Quincunx & X4. In a rather odd twist the
first 2 modes are RGSS, while the last mode is OGSS. Beneath
are some screenshots of each FSAA mode (& none) in
action in Return to Castle Wolfenstein.
No
FSAA
|
X2
FSAA
|
|
|
|
Quincunx
FSAA
|
X4
FSAA
|
|
|
|
As you
can see the results are fairly ok, although Quincunx is
rather blurry. That said the Voodoo 5 running at X4 FSAA still
appears better overall (At least based on my experience with
it).
Perhaps
in the future NVIDIA, much like 3dfx did, will release (Or
leak more likely) a Driver which has some manner of LOD
Biasing slider to offset the Texture blurring that can be
caused when using Quincunx & to a lesser degree the
other FSAA modes.
The
GeForce 3 features Hidden Surface Removal support, a feature
that's been used in PowerVR's Kyro 1/2 range of Graphics
Cards for quite some time now. Up until now none of the
traditional 3D Rendering Graphics Cards have used this
feature in Hardware though, although the ATI Radeon's Hyper
Z technology was to some extent the first implementation,
then 3dfx had a rather limited, BETA, CPU powered version
available in certain OpenGL Games (Quake 3 1.17 basically).
Although as discussed in another Guide, it involved the use
of a good few workarounds to get working without causing
much Rendering artefacts or other problems.
On
the GeForce 3 HSR is referred to as Z-Occlusion Culling.
This is a strangely odd name to use given NVIDIA’s liking
for marketable terms, e.g. HRAA, T&L, GPU, APU, etc.
Z-Occlusion Culling in comparison doesn't exactly roll off
the tongue & ZOC would just be stupid. Perhaps this is
because of the Power VR Graphics Cards out there which
already use this & NVIDIA's need to disassociate them
with using a similar feature.
HSR
is performed via the Lightspeed Memory Architecture using 2
methods to perform the HSR. The first being the depth test,
Z-Buffer values are compared to determine what is visible in
a scene, i.e. pixels at the front of the screen will obscure
pixels with a greater depth, behind them, e.g. If your
viewpoint is looking against a wall, then any pixels behind
that wall won't be visible through it. As such any
non-visible scene elements are completely disregarded, only
visible elements get sent onto the Framebuffer.
The
second is the Occlusion Query. This tests whether triangles
are hidden completely behind other triangles. As before,
non-visible triangles are disregarded & play no further
part in the rendering process.
The
benefits of HSR are twofold. 1. Is that it allows for more
efficient use of available Memory bandwidth. 2. Performance
improvements offered by rendering only what is visible. That
said this HSR is reliant on the rendering order of a scene
(Front to back only), in Games that have a more random
rendering order it’s essentially useless. Reportedly the
ATI Radeon 8500 (& perhaps the original too?) isn’t
rendering order dependent. Later on I’ll take a look at
the performance of HSR on the GeForce 3 Titanium 200.
|
