FAQ News Search Archive Forum IRC nZone Articles Tweaks Technology Files Prices Folding
Need Memory?
Popular Articles

GeForce3 Titanium Preview
By: Mike Chambers - October 1, 2001

Testing Configuration

In order to provide a broad measure of performance for the GeForce3 Ti 500, a variety of OpenGL and Direct3D based game and synthetic benchmarks were used. I had planned on testing with AquaNox, but received an out of memory error, which is the result of a known driver problem.

OpenGL Based:

  • Quake 3 Arena - version 1.29h
  • Return To Castle Wolfenstein Test - version 0.7.15
  • Serious Sam - version 1.02

Direct3D Based:

  • 3DMark2001 - build 200
  • Unreal Tournament - version 436
  • Giants - build 1.400
  • Motocross Madness 2

Non-antialiased performance was measured at resolutions of 1024x768, 1280x1024, and 1600x1200. Antialiasing performance was conducted at various resolutions using methods specific to each class of graphics card. Multiple graphics configurations were used in Quake 3 with each one offering improved visual quality. All tests were done in 32-bit color, with the exception of Motocross Madness 2, which was in 16-bit color. Results are reported in average frames per second.

The hardware used in this preview consists of the following:

  • Intel Pentium 3 - 1.0GHz @ 1.05GHz
  • 384MB Mushkin PC150 RAM (CAS 2-2-2)
  • Asus CUSL2 Motherboard (Intel 815E Chipset)

  • Detonator XP Driver Version 21.85
  • Sound Disabled / Vsync Disabled / 75Hz Refresh Rate
  • Windows 98 / DirectX 8.0

  • NVIDIA Reference GeForce3 Ti 500 (240MHz/500MHz)
  • Elsa Gladiac 920 GeForce3 (200MHz/460MHz)
  • NVIDIA Reference GeForce2 Ultra (250MHz/460MHz)

The motherboard's front side bus speed was increased from 133MHz to 140MHz. This provided a 50MHz increase in CPU speed and also allowed testing of all graphics cards on a slightly overclocked AGP bus.

I also received an updated Detonator XP driver (version 21.85) from NVIDIA. One of the updates allowed the device ID of the GeForce3 Titanium 500 to be recognized during driver installation.

Quake 3 Performance

It's come time to retire benchmarking with version 1.17 of Quake 3 and move on to what had been the latest version, which was 1.29h. However, during the course of writing this preview, version 1.30 was finally released. Version 1.29h comes with a new demo named FOUR and is a multiplayer match from The Campgrounds.

The Campgrounds

Multiple graphics configurations were used to test performance in Quake 3. Starting with the standard high quality settings, various options were enabled in subsequent settings that enhance image quality. Note that the default setting for texture compression (r_ext_compressed_textures) is 0, or off, in version 1.29h of Quake 3. The GeForce2 Ultra was only tested under settings 1, 2, and 3 since 32 and 64 sample anisotropic filtering, which are used in settings 4 and 5, are only supported on the GeForce3.

Quake 3 Test Settings

  Test Settings
1 Standard High Quality Settings
2 Setting 1 with High Geometry and Maximum Textures
3 Setting 2 with 16 Sample Anisotropic Filtering
4 Setting 2 with 32 Sample Anisotropic Filtering
5 Setting 2 with 64 Sample Anisotropic Filtering

Test Setting 1

Click to Enlarge - 218KB

Click to Enlarge - 218KB

Before enabling advanced texture filtering, or other visual improvements offered by Quake 3, certain graphic elements under the standard high quality settings have become rather bland.

Test Setting 5

Click to Enlarge - 248KB

Click to Enlarge - 248KB

Upon seeing the improvements offered by anisotropic texture filtering on the GeForce3, using bilinear or trilinear filtering alone is becoming a feature of the past. The above image is an example of 64 sample anisotropic filtering combined with trilinear texture filtering.

Demo FOUR - Setting 1

Demo FOUR - Setting 2

Looking at the GeForce3 Ti 500, we see marginal increases in performance over the GeForce3 under the high quality setting at 1024x768 and 1280x1024. As the graphics complexity increases in setting 2 with the addition of high geometry and maximum texture detail, so did the Ti 500's lead as it achieved a 10% increase in frame rates at 1280x1024 and 15% at 1600x1200.

What's most impressive with these results is that the GeForce3 Ti 500 manages to get 100 frames per second at 1600x1200 in setting 2 which is almost twice as fast as the GeForce2 Ultra which comes in at 54.6 frames per second.

Demo FOUR - Setting 3

With 16 sample anisotropic filtering in conjunction with trilinear filtering, the GeForce2 Ultra pulls in a respectable 103.4 fps at 1024x768. On the GeForce3 Ti 500, this same setting can be enabled at 1280x1024 and provide better performance (106.4 fps).

Demo FOUR - Setting 4

Demo FOUR - Setting 5

With 32 and 64 sample anisotropic filtering enabled, the GeForce3 Ti 500 offered increases in performance ranging from 12-19% over the GeForce3. With these settings, the increased performance isn't enough to justify a move up to the next highest resolution, but enabling a higher level of anisotropic filtering is possible. For example, moving from 32 sample anisotropic filtering at 1280x1024 on the GeForce3 (74.5 fps) to 64 samples on the GeForce3 Ti 500 (76.2 fps).

Next Page: Serious Sam & Wolfenstein Performance

Skip To:


Last Updated on October 1, 2001

Copyright 1998-2003. All rights reserved.
Reproduction in any form or medium without written permission of the site's owners is prohibited.

Privacy Policy

FastCounter by bCentral

 Visitors Are Online
Powered by Perlonline.com
Graphics Utilities
Add-In Cards
For Developers