Connect3D Radeon 9700 Pro in the test: performance leap at high resolutions

Connect3D Radeon 9700 Pro in the test: performance leap at high resolutions

Table of contents
  1. 1 Introduction
  2. 2 The card
  3. Scope of delivery
  4. 3 Drivers
  5. 4 Specs and technology
    1. 5 FSAA and AF
  6. 6 Benchmarks
  7. Test system
  8. 7 Synthetic tests
    1. 3DMark2001 SE
    2. 8 3DMark Detail
    3. 9 Villagemark D3D
    4. 10 Villagemark Detail
    5. 11 TempleMark D3D
    6. TempleMark D3D detail
    7. 12 Codecreatures Bench
    8. 13 Codecreatures Detail
    9. 14 Aquamark D3D
    10. 15 Aquamark D3D detail
  9. 16 Game benchmarks
    1. Ultima IX D3D
    2. 17 Ultim a IX D3D Detail
    3. 18 Comanche4 D3D
    4. 19 Comanche4 D3D Detail
    5. 20 Dungeon Siege
    6. 21 Dungeon Siege Detail
    7. 22 Aquanox
    8. 23 Aquanox in detail
    9. 24 Jedi Knight II
    10. 25 Jedi Knight II Detail
    11. 26 Max Payne
    12. 27 Max Payne detail
    13. 28 Serious Sam SE D3D
    14. 29 SeSam SE D3D Detail
    15. 30 Serious Sam SE OGL
    16. 31 SeSam SE OGL Detail
    17. 32 UT2003 Demo FlyBy
    18. UT2003 Demo Botmatch
    19. 33 UT2003 FlyBy Detail
    20. UT2003 Botmatch Detail
    21. 34 Alice
    22. 35 Alice Detail
  10. 36 Conclusion

Synthetic tests

Synthetic tests are often quite optimistic graphic demonstrations of how the manufacturer of the respective engine envisions the future of games. They can therefore give an outlook on future games and should therefore not be underestimated as an indicator. For the current situation at the gaming market, however, they are usually not too relevant in practice, as the benchmarks often show the maximum that can be extracted from a particular engine. In contrast to real games, this can be done safely because you want to advertise 'your' product and it does not depend on its suitability for the mass market, i.e. high sales figures.

3DMark2001 SE

3DMark2001 SE D3D v330 -default
Unit: points
  • 1024x768x32:
    • R9700pro (default)
    • GF4 Ti4600 (default)
  • 1280x1024x32:
    • R9700pro (default)
    • GF4 Ti4600 (default)
  • 1600x1200x32:
    • R9700pro(default)
    • GF4 Ti4600 (default)

From a good 2000 3DMarks in 1024 (approx. 18%) up to more than 2500 3DMarks in 1600, which is expressed by the lower absolute values ​​in a lead of 33.5%, the Radeon9700pro from Connect3D in our first benchmark can already achieve the GeForce4 Ti4600, the undisputed King of the Hill to date, very, very clearly. Since, as we have already seen in the technical data, the nominal raw output does not differ very clearly with regard to the pure texel fill rate, there must be other reasons for this. The following table provides information about a few interesting details.

3DMark2001 SE individual tests
ATiR9700 Pro nVidiaGF4 Ti4600 difference in 3DMarks 3DMarks 13590 11536 +2054 Game1 - Low Quality 185.1 183.6 +15 Game1 - High Quality 59.7 62.2 -50 Game2 - Low Quality 258.6 196.9 +617 Game2 - High Quality 139.9 115.2 +494 Game3 - Low Quality 173.2 166.2 +70 Game3 - High Quality 73.9 74.6 -14 Game4 - (High Quality) 97.6 51.4 +924 fill rate (single texturing) 1764.0 1092.2 fill rate (multi-texturing ) 2527.6 2331.9 High PolyCount (1 L.) 69.2 52.1 High PolyCount (8 L.) 15.0 12.6 Environment Bump Mapping 181.6 142.8 Dot3-Bump Mapping 192.9 154.1 Vertex Shader 171.6 96 Pixel Shader 170.1 124 Advanced Pixel Shader 189.4 97.1 Point Sprites 36.830,8

Except for the rounding error of 2 3DMarks, you can see quite clearly where the Radeon gets its points and thus also which tests are CPU-limited and which of the graphics cards have room for improvement grant.

As we did in our CPU scaling test with the Radeon9700 and the Ti4600, the Radeon9700 seems to need a little more CPU power as a tribute for its brute performance in order to really get going, so that the Ti4600 can keep up very well in purely CPU-limited scenarios and even two in this low resolution who can decide for themselves. However, as soon as the resolution is increased, the Geforce4 must also pass the test Game3- High Quality and in 1600x1200 also Game1 - High Quality to the Radeon9700pro.

The performance in the triangular throughput is also interesting. With its four vertex pipelines, the Radeon9700pro should theoretically achieve a triangular power twice as high as the GeForce4 with its two vertex shaders. Converted to the same clock rate, this only works at least partially in the vertex shader test with a lead of 65%, whereas the high polygon count tests only attest the Radeon9700pro a lead of 10% to a maximum of 22.5%. Even if this test is carried out with DirectX7 methods, which Madonion unfortunately does not comment on, the Radeon9700pro should be significantly faster, unless we are already dealing with a CPU limitation. In resolutions lower than 1024, the GeForce4 can catch up with the Radeon9700pro in the tests Point Sprites, Pixel Shader and Dot3-Bump Mapping. This suggests that these functions can be performed faster on the Radeon9700pro than on the GeForce4 Ti mainly due to the gigantic memory bandwidth and its more effective use.

Let's now take a look at the performance of bothCards when, as in gaming, the abundant raw performance is not wasted in senseless 150fps and more, but invested in anti-aliasing and higher-quality texture filtering.

3DMark2001 SE D3D v330 - Quality
Unit: Points
  • 1024x768x32:
    • R9700pro (2xAA/4xAF)
    • R9700pro (4xAA/8xAF)
    • R9700pro (6xAA/16xAF)
    • GF4 Ti4600 (2xAA/4xAF)
    • GF4 Ti4600 (4xAA/8xAF)
    • GF4 Ti4600 (4xS-AA/8xAF)
  • 1280x1024x32:
    • R9700pro (2xAA/4xAF)
    • R9700pro (4xAA/8xAF)
    • R9700pro (6xAA/16xAF)
    • GF4 Ti4600 (2xAA/4xAF)
    • GF4 Ti4600 (4xAA/8xAF)
    • GF4 Ti4600 (4xS-AA/8xAF)
  • 1600x1200x32:
    • R9700pro (2xAA/4xAF)
    • R9700pro (4xAA/8xAF)
    • R9700pro (6xAA/16xAF)
    • GF4 Ti4600 (2xAA/4xAF)
    • GF4 Ti4600 (4xAA/8xAF)
    • GF4 Ti4600 (4xS-AA/8xAF)

A result that couldn't be more clear! The Radeon9700pro wipes the floor with the GeForce4 as soon as you use the quality settings. The almost four and a half thousand points that the GeForce4 still achieved with the medium settings 4xFSAA and 8xAF in 1024 are not a continuous jerk orgy, but you might expect a bit more from a high-end card than just average performance. The Radeon can offer this, even with significantly better anti-aliasing and a higher set anisotropic filter, all tests of the 3DMark2001 run almost smoothly up to the 1280 resolution. Only in 1600 with 4x/8X settings or higher could occasional jerks spoil the still impressive performance a little.

Since the Radeon9700pro leaves no doubt as to what it was developed for, we can only see on the next page , which is the reason for the massive drop in performance of the GeForce4: the FSAA or AF.

On the next page: 3DMark Detail