GeForce3 Titanium Preview
By: Mike Chambers - October 1, 2001
3DMark2001 Performance
Many gamers enjoy testing their 3D graphics performance with MadOnion's 3DMark2001 since it can be downloaded for free and is straightforward to use. They also provide a web-based service where scores can be saved and compared to other systems. Although it's referred to as a synthetic benchmak, the battery of tests in 3DMark2001 contains a variety of state-of-the-art simulations, which could easily consist of what a Direct3D game is based on. The default settings were used for these tests which consist of 32-bit color, double frame buffer, compressed textures, and 24-bit z-buffer running at a resolution of 1024x768.
I've run across comments in regards to image quality with initial release of the Detonator XP drivers and 3DMark2001. In the Game 1 test, smoke trails are missing from the rockets while in the Game 2 test, the quality of the fog has slightly degraded. None of these conditions occurred on my system using an updated Detonator XP driver (21.85).
Game 1
Game 3
Nature
I still can't get over how awesome the Nature demo is...
3DMark2001 - Overall Score
The following table shows the individual game test results, which are used to determine the final 3DMark2001 score. Each test consists of a low quality (LQ) and high quality (HQ) setting. Note that the Nature test does not run on graphics cards, such as the GeForce2 since they don't support pixel shaders which cannot be emulated in software mode.
3DMark2001 - Individual Test Results
Test
GeForce3 Ti 500
Elsa GeForce3
GeForce2 Ultra
Game 1 - LQ
94.3
90.6
73.7
Game 1 - HQ
29.8
29.8
22.3
Game 2 - LQ
96.0
86.3
57.8
Game 2 - HQ
52.9
47.7
21.7
Game 3 - LQ
99.0
96.4
75.5
Game 3 - HQ
45.0
44.8
35.1
Nature Scene
35.1
32.3
N/A
DirectX 8 compliant graphics cards, such as those based on the GeForce3, perform very well in 3DMark2001. On the GeForce2 certain graphics operations which execute extremely fast on the graphics processing unit, are offloaded to the central processing unit. In these situations a faster processor will provide better performance.
More Direct3D Tests
I also tested performance in a variety of Direct3D based games which included Unreal Tournament (Thunder demo), Giants (gamegauge), and Motocross Madness 2 (self recorded demo). Graphics settings for Unreal Tournament and Giants were set to maximum levels while the default settings were used in Motocross Madness 2.
Motocross Madness 2
Motocross Madness 2 was released in May of 2000 by Microsoft. If you're into racing simulations, I highly recommend this game.
GeForce3 Ti 500 Direct3D Performance
Game
GeForce3 Ti 500
Elsa GeForce3
GeForce2 Ultra
Unreal Tournament
800x600
99.1
99.0
99.1
1024x768
95.2
94.9
95.7
1280x960
94.0
92.8
84.6
Giants
1024x768
56.4
56.6
53.8
1280x1024
55.6
54.8
46.8
1600x1200
52.7
51.3
36.4
Moto. Madness 2
1024x768
57.0
57.0
55.0
1280x1024
52.0
49.0
45.0
1600x1200
45.0
41.1
37.0
Both GeForce3 cards are running pretty even until 1600x1200 where the GeForce3 Ti 500 was able to squeeze out a few extra frames in all the games tested. However the most startling discovery in these results was gaining an additional 14 frames per second in the Unreal Tournament Thunder demo when compared to my 1GHz Pentium 3 review which was done using the Detonator 3 12.90 drivers.
Keep in mind that performance from the Giants gameguage benchmark is not indicative of actual gameplay as it's based on a movie mode. Gameplay performance is normally lower.
