Gainward Ultra/1100XT TV-DVI Video CardTue, Apr 27, 2004 - 9:00 AM
- Intel 2.4C
- ABIT AI7
- 1Gb Kingston PC3500 RAM (dual channel)
- Maxtor/Quantum Fireball 40Gb HDD
- Thermaltake PurePower 360W PSU
- Gainward Ultra/1100XT TV-DVI 128mb (Default is 450/780)
- Albatron FX5900 XTV 128mb (Default is 390/700)
- ASUS A9600XT/TVD 128mb (Default is 500/600)
* ASUS 9600XT compared to the Gainward Ultra/1100XT. Note the length difference.
Our test setup includes Windows XP SP1a, DirectX 9.0b, Forceware 56.72 driver for the 5900XTs and Cat 4.4 driver for the 9600XT. All benchmarks were completed with no other software programs running. For overclocking the Gainward Ultra/1100XT TV-DVI, we used the bundled EXPERTool Utility. The desktop resolution was set to 1024x768x32. A brief description of each test will be provided at the beginning of each test section. Each test was run 3 times in a row and the average of the tests is the result shown in our graphs. In the game tests, FRAPS was used to acquire the data points used in the line graphs, while the FPS score was taken from each game’s timedemo results. We ran the following tests:
- Synthetic Benchmarks
- 3dMark2001SE build 330
- ShaderMark v2.0 build 1e
- D3D FableMark v1.0
- Fillrate Tester
- Spec Viewperf 7.1.1
- UT2003 v2225
- Call of Duty
- Call of Duty
- 3dMark2001SE build 330
3dMark2001SE build 330
3dMark2001SE was run at default setting – 1024×768 and NoAA (shown in the graph), along with 1024×768 and 4xAA, 1280×1024 and NoAA, and 1280×1024 and 4xAA. AA was set through the software.
Here we see that the two 5900XT’s run very close to each other in performance. Take note that it does not seem that the extra clock speeds on the Gainward card show any significant performance boost.
The same lighting issues in the Nature game test exist on the Gainward card as they did on the Albatron card. The sunlight is so bright that it starts taking away from the texture of the tree leaves. This issue is nothing that Gainward or Albatron could control because it is pretty much known that this issue is an architectural issue with the GeForce FX core technology.
AquaMark3 was run at 1024×768 with NoAA and 4xAA set through the driver.
Unlike 3dMark2001SE, we do see the clock speed differences in AquaMark, but only for the NoAA test. Adding 4xAA through the driver brings the two 5900XT cards to the same score. You can see that the 9600XT performs a little behind the 5900XTs in this test.
ShaderMark v2.0 build 1e
ShaderMark was run at 1024×768 with NoAA.
When we ran ShaderMark, we got a series of warnings about how certain shaders were not supported by our current hardware. Those shaders that were not supported received no score. What is very interesting about this test was that the 9600XT showed a much higher score than 5900XTs. This may relate to the image quality difference between the GeForce FX and ATI’s Radeon series of cards.
D3D FableMark v1.0
FableMark is a quick little benchmark that is great for testing stencil edges. We tested the 9600XT’s at 1024×768 with NoAA and 4xAA/8xAF set through the driver.
Here Gainward’s card pulls above Albatron’s. This gain can be related to the clock speed enhancements present for the Gainward card.
Fillrate Tester was run using Display Mode 1024×768 @ 85Hz. No other changes to the settings were done.
I really like this test. The results in this test verify the results shown in ShaderMark, where the 9600XT scores much higher on the shader-related tests. Also present in these results is the clock speed benefits of the Gainward card over the Albatron.
Spec Viewperf 7.1.1
Spec Viewperf stresses the OpenGL side of the cards. We ran all the tests at default settings.
The results from Spec Viewperf show no real advantage for the Gainward clock speeds over the Albatron. It does show the horse-power advantage of the 5900XTs over the 9600XT, except in the ‘ugs-03′ test.