Date:
Sunday , July 20, 2003
Author:
Brent Justice
Editor:
Kyle Bennett
Google +1

UT2003 Texture Filtering

ATI recently questioned the validity of our UT2K3 benchmarks in our BFGTech 5900 Ultra review. ATI explained to us that our benchmarks were damaging to their 9800 series card. We look further into this and share the results with you.

Introduction:

Editor's Note: As mentioned above, ATI has come to us and told us that our UT2K3 benchmarks were off in our BFGTech Asylum 5900 Ultra review and that the benchmark numbers we shared were damaging the 9800 line of products. This was first brought to me by their PR Director, Chris Evenden, last Tuesday morning. At that time they did not submit any proof of their claims that shed light on the specific benchmarks we used. As of posting this, ATI has still never given us proof of any of their statements specific to the benchmarks they suggest are damaging to their product. Still, there was no doubt that ATI was genuinely concerned with the benchmark numbers. HardOCP has been incredibly critical of Image Quality this year and we have not been afraid to share our results with you, as was shown here in March of this year in this article that directly called NVIDIA's IQ into question. With that said, we are genuinely concerned with representing video cards performance to our readers fairly. So with ATI's prompting, Brent Justice and I decided to look further into the matter, but did so independently. Meaning that he did his own testing, as did I, and this article represents the joint conclusions we arrived at.

This article focuses on the quality of filtering the GeForceFX series of video cards are producing in Unreal Tournament 2003 with 44.03 drivers. Specifically Trilinear Filtering but touches on Anisotropic Filtering as well.

We have found that the main issue stems from the fact that the 44.03 drivers do not have an “Application Preference” option like the ATI control panel does. By selecting “Application Preference” in the ATI control panel you are allowing the driver to use whatever filtering the application specifies, or rather what the game developer has decided it should filter. With the current 44.03 drivers from NVIDIA we have three quality settings, “High Performance", “Performance” and “Quality”. You might think that when you set the driver to “Quality” that it would use the best filtering known as full Trilinear Filtering, although the 44.03 driver from NVIDIA does not seem to allow this...but does that matter?

Editor's Note: Trilinear Filtering is best described as how textures seen in your game transition between each other as they fall further away from your viewpoint. Say if you are looking down a hallway with a tile floor. The game engine and drivers will make the tile texture less detailed the further away it is and does this in blocks, or mipmaps. Say you have three visible blocks the second is less detailed than the first and the third less detailed than the second. Instead of leaving a very visible line in between the blocks, video cards do filtering between the blocks to make the transition less noticeable. With Bilinear Filtering in games you might notice "a line" somewhere towards the horizon especially as you move. Trilinear Filtering is a technique to get rid of that line, but can be done in several different ways.

We are going to examine if there is a difference in filtering and if this difference actually decreases in-game image quality and gameplay experience specifically in UT2K3 comparing ATI's 9800 and NVIDIA's 5900 series chipsets.

Test Setup:

ABIT IC7-G (i875P), Intel Pentium 4 3GHz “C” operating at 800MHz FSB, 2 X 512MB Corsair XMS PC3200LL TwinX Dual Channel DDR400, Maxtor 40GB ATA/133, Windows XP Professional SP1 with DirectX 9.0a.

BFG Asylum 5900 Ultra 256MB – Operating at default speeds and using 44.03 drivers.

The NVIDIA driver control panel was set to the image quality setting of “Quality” with AA disabled. Anisotropic filtering was set to OFF for the screenshots with No AF and then it was set to 8X AF when we tested with AF.

ATI Radeon 9800 Pro 128MB – Operating at default speeds and using Catalyst 3.6 suite drivers.

The ATI driver control panel was set to “Application Preference” on both AA and AF so that we could test with AF disabled. Then we set the AF slider to Quality AF when AF was tested.

All screenshots were taken with the latest version of HyperSnapDX and verified that they do represent what was seen in-game.

All screenshots were taken at 1280x1024. No alterations were made beyond splicing the images. They were saved in a .PNG file format from their original BMP format. Then they were reduced in size using Ulead Smart Saver Pro in JPEG format at "Q95 C95 411 Progressive" quality. The file sizes of the pictures are still quite large. The largest file is 451KB. We are sorry if this not easily usable for our modem readers, but it is simply impossible to discuss the levels of IQ that we address here using images that are compressed more than what we already have to work with.

For all the true DirectX geeks out there, we will list the relevant parts of the UT2003.INI file that deal with the 3D settings used in the game.

[WinDrv.WindowsClient]

WindowedViewportX=640

WindowedViewportY=480

FullscreenViewportX=1280

FullscreenViewportY=1024

MenuViewportX=640

MenuViewportY=480

Brightness=1.000000

Contrast=0.700000

Gamma=0.800000

UseJoystick=False

CaptureMouse=True

StartupFullscreen=True

ScreenFlashes=True

NoLighting=False

MinDesiredFrameRate=0.000000

Decals=True

Coronas=True

DecoLayers=True

Projectors=True

NoDynamicLights=False

ReportDynamicUploads=False

TextureDetailInterface=UltraHigh

TextureDetailTerrain=UltraHigh

TextureDetailWeaponSkin=UltraHigh

TextureDetailPlayerSkin=UltraHigh

TextureDetailWorld=UltraHigh

TextureDetailRenderMap=UltraHigh

TextureDetailLightmap=UltraHigh

TextureMaxLOD=0

TextureMinLOD=0

NoFractalAnim=False

ScaleHUDX=0.000000

MouseYMultiplier=0.000000

MouseXMultiplier=0.000000

[D3DDrv.D3DRenderDevice]

DetailTextures=True

HighDetailActors=True

SuperHighDetailActors=True

UsePrecaching=True

UseTrilinear=True

AdapterNumber=-1

ReduceMouseLag=True

UseTripleBuffering=False

UseHardwareTL=True

UseHardwareVS=True

UseCubemaps=True

DesiredRefreshRate=85

UseCompressedLightmaps=False

UseStencil=False

Use16bit=False

Use16bitTextures=False

MaxPixelShaderVersion=255

UseVSync=False

LevelOfAnisotropy=1

DetailTexMipBias=0

DefaultTexMipBias=0

UseNPatches=False

TesselationFactor=1.0

CheckForOverflow=False

AvoidHitches=False

OverrideDesktopRefreshRate=False

If you wish to try your own testing you may download the UT2003.INI file that we used for these image quality tests. Copy this file into the System folder of your UT2003 folder. Don’t forget to backup your original UT2003.INI file first though. Once you place it in your system folder of your UT2003 directory and rename it you may simply run the game and you will have the same game settings we used for this article and our benchmarking. The resolution is set for 1280x1024 with an 85Hz refresh rate, so if your monitor doesn’t support that resolution and refresh rate it may not work, you may need to lower the refresh rate in the INI file first.