Part 4: Rendering Quality, New Anti-Aliasing and Anisotropic Filtering Modes
(Continuation, Half-Life 2)

TABLE OF CONTENTS

1. Part 1. Theory and architecture
2. Part 2: Practical examination
   • Video cards' features
   • Testbed configurations, benchmarks
   • Synthetic test results
3. Part 3: Game test results (performance)
4. Part 4: Render quality comparison in games
   • Synthetics: D3DAFTester and DX9FSAAViewer
   • Far Cry
   • Doom 3
   • Half-Life 2
   • Splinter Cell: Chaos Theory
   • F.E.A.R. SP Demo
   • Serious Sam 2 Demo
   • Conclusions on Part IV

Settings and conventional signs

• AA and AF control was done from the game.
• Drivers were configured by default, except for those modified to analyze the quality in various AA and AF modes.

• 4x/16x = MSAA 4x, AF 16x
• 4x+TA/16x = MSAA 4x + Transparency/Adaptive AA, AF 16x
• 4x+TA/16x+HQ = MSAA 4x + Transparency/Adaptive AA, AF 16x High Quality (HQ AF and Catalyst A.I. Off in R520, Catalyst A.I. Off in R480 and HQ mode in G70)

Half-Life 2

The third game, where we shall compare rendering quality of the three video cards, is Half-Life 2. The first official sequel to the game is about to be released — Half-Life 2 Aftermath.

	Radeon X1800	GeForce 7800
4x/16x	JPEG (~530 KB) | PNG (~2.7 MB)	JPEG (~530 KB) | PNG (~2.7 MB)
4x+TA/16x	JPEG (~530 KB) | PNG (~2.7 MB)	JPEG (~530 KB) | PNG (~2.7 MB)
4x+TA/16x+HQ	JPEG (~530 KB) | PNG (~2.7 MB)	JPEG (~530 KB) | PNG (~2.7 MB)

	Radeon X850
4x/16x	JPEG (~530 KB) | PNG (~2.7 MB)
4x+TA/16x	-
4x+TA/16x+HQ	JPEG (~530 KB) | PNG (~2.7 MB)

Of course, the difference in antialiasing quality of wires is striking: that's the effect of no effective gamma correction in the GF7800. Then, a prying glance makes out a difference in antialiasing quality of transparent textures — trees. It's hard to tell the reasons for this difference. But if you take a closer look, it bears down on you that it's also present on the screenshot without TAA/AAA. To all appearances, we again see the effect of different MIP mapping algorithms. Another possibility: just a different implementation of rendering transparent textures in the game code for GF and R.

The next evident difference is the low quality of texture filtering in GF7800 (we have already seen it in Doom 3) — these fail-places are especially noticeable on the stone pavement in the left part of the screenshots. The cure is the same: as soon as you switch ForceWare to High Quality, the filtering quality immediately gets identical to that in Radeon.

Speaking of Radeon. That's the third game already, where we don't see significant advantages of the new anisotropy algorithm in the RX1800. Strictly speaking, there are no insignificant advantages either. The dull picture is brightened up only by the high performance of the HQ anisotropy. That is, no advantages is OK, if the performance is practically the same.

Back: Doom 3 | Forward: SC: Chaos Theory