Creative Labs 3d Blaster GeForce3 Ti500

Author : Wayne Date : 24th September 2001

3DVelocity would like to thank Creative Labs and especially Rosie Tickner of ProdigyPR for their help and courtesy in providing this graphics card for review.

Benchmarks :

DroneZ Mark :

I ran this benchmark using the "Generic High Quality" setting to maintain compatibility with future tests run on other cards.

	Minimum	Maximum	Average
FPS	56.55	340.85	127.86
T+L K Triangles	14.35	3431.62	1874.39
OpenGL K Triangles	13.59	3431.27	1855.81

DroneZ Mark is a feature rich OpenGL Benchmark which I should point out is optimised for the GeForce and also for Intel processors. It uses a lot of OpenGL visual tricks to push your GPU to its limits though as you can see from the 340fps maximum scored here, not all scenes are equally demanding. The Radeon ran this demo quite well in the end, but had some real problems in the opening menus where it would stutter and freeze for seconds at a time. Again the 8500 gets well and truly hammered by the results, though this is perhaps less of a defeat when we allow for the GeForce optimisations.

 

Final Reality :

This may be an old benchmark, but it's great for checking out retro-compatibility. It uses only MMX and AGP optimisations and relies heavily on pure muscle.

Another solid performance with the highest 2D score to date. 3D performance and bus transfer came in fractionally behind the Radeon 8500.

NBench :

AMD's NBench isn't particularly impressive to watch, not unless you're heavily into the whole Anime, Manga type scene anyway. What is does do though is give your graphics card and CPU a thorough pounding.

VillageMark

VillageMark is designed to test a graphics card's ability to handle overdraw (hidden surface removal). A flying camera pans around a village rendered in large (up to 1024x1024) high quality textures. The fact that they are constantly appearing and being hidden as the camera pans makes heavy use of your GPU's ability to effectively discard unwanted data before it's rendered then dumped.

This is less of a head turner for the GeForce suggesting they still have some work to do on their Z-occlusion culling routines. Remarkably the Radeon 8500 managed an awesome 94fps in this test, turning out 63fps with its HyperZ II function disabled!

FSAA Performance :

Just to get a feel for the performance hit imposed by using FSAA, I ran the Q3A timedemo in each of the three AA modes and compared the results to the non AA scores. The demo was run using maximum quality settings and at the most likely resolutions to benefit from FSAA, those being 800x600 and 1024x768.

The idea of running Q3A at 10x7 with 4 sample AA at 77 frames per second is enough to get even the most hardened gamer drooling. It should be immediately obvious that even under the most demanding of situations, NVIDIA's high resolution anti-aliasing is at last making this a feasible technology rather than a feature on the box that's barely worth enabling.

Just as a double check I re-ran the same tests using Vulpine at 1024x768 using the standard OpenGL 1.2 feature set, high detail and 32 bit textures/resolution. Here's the breakdown :

Quincunx certainly offers a significant performance advantage over 4xAA, and despite the slight lack of definition in the textures it's a good trade-off. You may find you prefer 4x or even 2x in certain games rather than Quincunx but it's great to have the ability to choose, and I personally don't find the slight softening of the textures unpleasant.

For those of you who have sat through Vulpine's GLMark with AA switched off, let's just say that if you admired the curves on the chick from the helicopter that way, seeing them in NVIDIA's glorious 4sample AA at high resolution borders on voyeuristic.

Click for a larger image

Page10 - Overclocking