|
System Specifications
| CPU: |
Intel Pentium IV 2.4C (12*200MHz)
|
| Motherboard: |
MSI 865PE Neo2-FISR
|
Abit IS7 Max II Advance |
| Memory: |
2*512MB Corsair TwinX4000 |
2*256MB Corsair TwinX3200 |
| Hard Drives: |
40GB Seagate Barracuda IV 7200RPM, 2*80GB Maxtor SATA 7200RPM
|
80GB Western Digital HDD
|
| Video Card: |
Matrox Parhelia 128MB (200MHz/250MHz) |
ATi AiW Radeon 9800 Pro 128MB (380MHz/340MHz) |
| |
MSI NBox FX5900 Ultra-VTD 256MB (450MHz/850MHz) - (500MHz/951MHz) |
MSI FX5900-VTD 128MB - (400MHz/850MHz) |
| |
ATi AiW Radeon 9000 Pro 64MB (275MHz/270MHz) |
|
| Operating System: |
Windows XP Professional Service Pack 1 w/Direct X 9 |
| Drivers: |
Parhelia 1.04.03.005 |
ATi AiW Radeon 9800 Pro - Catalyst 3.6 |
|
MSI NBox 5900 Ultra-VTD - 45.23 |
MSI FX5900-VTD 128MB - 44.03 |
|
ATi AiW Radeon 9000 Pro - Catalyst 3.7 |
|
| Cooler: |
Vantec AeroFlow |
| Case: |
CoolerGuys Windtunnel IV |
| Power Supply: |
RaidMax 400Watt Power Supply |
| Software: |
Fraps 2 |
| Direct X Benchmarks: |
Unreal Tournament 2003 (HardOCP 2.1 software Antalus Demo) |
Code Creatures |
| |
Gun Metal |
X2: The Threat |
| |
Splinter Cell (Beyond 3D Demo) |
|
| OpenGL Benchmarks: |
Jedi Knight II Jedi Outcast (time demo) |
| |
Serious Sam SE (Little Trouble) |
Spec Viewperf 7.0 |
As you can see from the table above there are more DirectX based benchmarks at the moment than that of OpenGL. However we have kept some of the OpenGL tests that matter, Jedi Knight II taking care of the Quake III engine games, Serious Sam SE with its own OpenGL implementation and Spec Viewperf in the professional area. For DirectX we have decided to take a look at two new DirectX 9 benchmarks, Gun Metal and X2: The Threat. This should give us a indication of DirectX 9 performance more so that other synthetic tests. The only major synthetic test that we have is Code Creatures which looks good but don't place much stock in its frame rates. Lastly we are using the common UT2003 benchmark in flyby mode, using the one of the most demanding maps, dm_Antalus.
UT2003, Splinter Cell, Jedi Knight II, and Serious Sam SE were all run at 1024*768 and 1600*1200. For all but the Parhelia there were also 4X AA and 8X ansiotropic filtering results included in these tests, while the Parhelia used 16X FAA and 2X ansiotropic filtering. Gun Metal and X2 were only tested at 1024*768 both with and without AA and ansiotropic filtering enabled. Code Creatures was only run with AA and ansiotropic filtering disabled using the standard full test. All games, minus UT2003 had Fraps running in the background to capture the actual fps for each second, saved to a text file.
There were two different systems used for testing, one is Huberts from his AiW 9800 Pro review, as the results are also from there. The other system, used to test the 5900Ultra, AiW Radeon 9000 Pro and the Parhelia was run using the latest drivers at the time of writing. So without further wait let us look at the synthetic tests. Synthetic Tests
Code Creatures is really a nature scene put though the DirectX 8.1 system, to create a high quality benchmark with a large amount of polygons. Depending on the card you can have over 22 million polygons on the screen at once. However it is still a synthetic benchmark, a program that you cannot play, so keeping this in mind how does the NBox 5900U do?
|
Card
|
1024*768 Avg FPS |
1280*1024 Avg FPS |
1600*1200 Avg FPS |
| ATi Radeon AiW 9800 Pro: |
47.7 |
39.5 |
31.3 |
| MSI FX5900U Oc'd: |
42 |
37.7 |
32.9 |
| MSI FX5900: |
42.1 |
33 |
26.3 |
| MSI FX5900U: |
40.9 |
35.8 |
30.3 |
| Matrox Parhelia: |
17.4 |
13.5 |
10.3 |
| ATi Radeon AiW 9000 Pro: |
14 |
9.6 |
7.6 |
We can see that for the lower resolution tests the Radeon 9800 is the best by a bit at least. It is only at 1600*1200 that the overclocked 5900 Ultra finally bests the 9800 Pro. If we look at the higher resolutions we see that the Ultra version of the 5900 is at least 2 to 4 fps (8-15%) faster than the regularly clocked version which is around the 12% difference in clock speed. When overclocked the Ultra gains about 2 fps in all resolutions, which boosts the 5900 Ultra above the Radeon 9800 here. What about when we look at a more professional use for these cards in Spec Viewperf.
SPECviewperf is the standard by which professional video cards are tested. The sole goal of the SPEC organization is to create and maintain a standard set of useful/relevent benchmarks. As such SPECviewperf is the graphics test, aimed at professional users, those who use 3D rendering, Pro Engineer as well as other high end programs. So how does this video card fair in this test, has nVidia provided a card that is capable of being used in a professional application?
| SPECviewperf 7.0 |
Parhelia |
MSI FX5900 Ultra |
ATi Radeon AiW 9800 |
MSI FX5900 Ultra Oc'd |
MSI FX5900 |
ATi Radeon AiW 9000 |
| 3dsmax-02 |
9.255 |
14.22 |
14.44 |
14.65 |
13.41 |
5.819 |
| Drv-09 |
19.51 |
52.36 |
39.07 |
53.62 |
48.06 |
31.3 |
| Dx-08 |
51.57 |
39.52 |
60.83 |
39.6 |
31.97 |
41.94 |
| Light-06 |
9.801 |
11.26 |
12.42 |
11.32 |
11.01 |
11.25 |
| ProE-02 |
4.364 |
15.19 |
15.48 |
15.8 |
14.93 |
10.02 |
| Ugs-03 |
6.286 |
7.717 |
24.99 |
8.519 |
6.74 |
5.975 |
The results we see from SPEC show that the FX5900 series does very well. The increase in clock speed nets at least some improvement in score with these tests. When overclocking the Ultra we see a small increase in speed, except in the Ugs-03 test where there is a 10% increase in the score. The GeForce FX cards do especially well in the Drv-09 test, besting the Radeon 9800 by a fair margin. However in the Dx-08 test we see that all the other cards best the FX series of cards, even the Matrox Parhelia which gives a score near that of the Radeon 9800, and the Radeon scores extremely well in the Ugs-03 test. Overall a good showing in this test for the Ultra with it performing better than the regular 5900 and getting close to that of the Radeon 9800 in almost all tests.
|
