3DMark 2001SE &3DMark 2003
This content was originally featured on Amdmb.com and has been converted to PC Perspective’s website. Some color changes and flaws may appear.
3DMark 2001SE remains one of the established methods of comparing performance of video hardware with each other. Here we will compare the results using this DirectX 8.1b based benchmark.
| Radeon 9700 | Radeon 9500 | FX5600 | FX5200 | Ti4800SE | |
| 3DMarks | 12211 | 10312 | 8966 | 7424 | 11251 |
| GT1 – Car Chase (Low)* | 163.03 | 147.13 | 129.77 | 103.50 | 164.43 |
| GT1 – Car Chase (High)* | 51.67 | 50.70 | 47.53 | 47.43 | 55.67 |
| GT2 – Dragolithic (Low)* | 238.87 | 183.43 | 136.33 | 110.23 | 199.83 |
| GT2 – Dragolithic (High)* | 126.43 | 103.50 | 74.23 | 61.93 | 111.27 |
| GT3 – Lobby (Low)* | 147.43 | 142.07 | 133.73 | 121.47 | 147.73 |
| GT3 – Lobby (High)* | 65.27 | 65.43 | 61.93 | 59.13 | 67.70 |
| GT4 – Nature* | 92.60 | 59.70 | 64.70 | 35.10 | 71.90 |
| Fill-Rate (Single) | 1545.20 | 918.13 | 838.37 | 627.80 | 967.73 |
| Fill-Rate (Multiple) | 2190.47 | 1105.00 | 1107.30 | 732.23 | 2129.57 |
| High Poly (1 light) | 55.50 | 52.23 | 40.23 | 31.33 | 49.70 |
| High Poly (8 light) | 14.00 | 13.80 | 7.40 | 5.00 | 11.50 |
| Environment Bump Mapping | 202.37 | 153.83 | 122.60 | 70.80 | 145.43 |
| DOT3 Bump Mapping | 186.30 | 113.63 | 99.40 | 73.60 | 138.17 |
| Vertex Shader | 156.93 | 153.27 | 64.93 | 46.47 | 91.07 |
| Pixel Shader | 224.57 | 162.37 | 143.43 | 87.47 | 115.80 |
| Adv. Pixel Shader | 181.53 | 95.40 | 50.47 | 30.10 | 93.60 |
| Point Sprites | 34.20 | 13.03 | 20.90 | 12.00 | 23.50 |
| . | |||||
| *Only GT1-GT4 are used in calculating the overall 3DMark score. | |||||
Looking at the results, we see that the Ti4800SE continues beating the FX5600 in every test (with the exception of UT2K3 with AF turned on). Both the Radeon 9500 and 9700 beat the rest in vertex shading and pixel shading – features used heavily by game developers today.
An interesting note is that the Ti4800SE does really well in the fill-rate department – considerably better than the FX5600. In terms of raw performance of filling the screen-buffer, the Ti4800SE is amazing (which is what a lot of fans like to see).
3DMark 2003
3DMark 2003 is the next step in Futuremark’s graphics benchmarking tool featuring some DirectX 9.0. 3DMark 2003, in my opinion, is not a very good measurement if you’re looking for DirectX 9 comparisons because only 1 of the 4 game tests actually features any DirectX 9 at all. (Game Test #4 – Mother Nature is the only test with some DirectX 9 implemented).
Also, if you follow the following formula for calculating 3DMark 2003 scores, you can see that the fraction of DirectX 9 in “GT4” below is only a very small amount compared to the rest of the equation (which is largely DirectX 8.1 and DirectX 7).
*Source: [H]ardOCP.
| Radeon 9700 | Radeon 9500 | FX5600 | FX5200 | Ti4800SE | |
| 3DMark Score | 4158 | 2805 | 2054 | 1226 | 1545 |
| GT1 – Wings of Fury | 128.9 | 103.5 | 91.0 | 67.4 | 100.1 |
| GT2 – Battle of Proxycon | 28.8 | 16.4 | 14.4 | 6.7 | 9.9 |
| GT3 – Troll’s Lair | 25.4 | 15.6 | 12.2 | 5.9 | 9.6 |
| GT4 – Mother Nature | 24.8 | 18.3 | 7.3 | 5.3 | N/A |
| CPU Score | 397 | 400 | 411 | 443 | 446 |
| CPU Test 1 | 42.2 | 42.5 | 41.4 | 44.5 | 45.4 |
| CPU Test 2 | 7.4 | 7.5 | 8.0 | 8.7 | 8.6 |
| Fill Rate (Single-Texturing) | 1308.5 | 847.5 | 787.5 | 605.1 | 740.8 |
| Fill Rate (Multi-Texturing) | 1988.1 | 1083.0 | 1056.3 | 707.0 | 1990.7 |
| Vertex Shader | 13.4 | 13.0 | 4.3 | 3.2 | 5.7 |
| Pixel Shader 2.0 | 37.5 | 24.0 | 6.0 | 4.3 | N/A |
| Ragtroll | 18.4 | 11.8 | 7.9 | 4.4 | 5.9 |
| No sounds | 28.5 | 28.9 | 27.5 | 28.9 | 29.5 |
| 24 sounds | 24.9 | 25.1 | 24.0 | 25.4 | 25.7 |
| 60 sounds | 21.5 | 21.8 | 21.0 | 22.5 | 22.2 |
We can see that the Radeon 9700 continues to beat the rest in benchmarks. What is more interesting are the numbers generated by the Ti4800SE. We can see that the non-“Game Test” results are very good and beats the FX5600 in practically each test run. Then why is it that the Ti4800SE is scoring below the FX5600?
The reason the FX5600 beats the Ti4800SE in this benchmark is because of PixelShader 1.4 (used in Game Tests #2, #3, and #4). The Ti4800SE having only support for PS 1.2 requires more processing to render the same scene when compared to the FX5600 (using PS 1.4). Also, without Pixel Shading 2.0 the Ti4800SE can’t run Game Test #4, and therefore scoring a 0 in the equation above.
You add up these crippling effects and you can see that the only test in which the Ti4800SE competes on equal footing is in Game Test #1 which is DirectX 7 and no pixel-shading. Therefore it’s no coincidence that the Ti4800SE beats the FX5600 in this test.
It is for this reason that 3DMark 2003 is unfair for older video cards not supporting Pixel Shading 2.0 and Pixel Shading 1.4. It’s like bringing a horse to a race and finding out everyone else is driving Formula 1 race cars. You get the idea. Actually, I would not have even tested the Ti4800SE with 3DMark 2003 if it weren’t for the fact I wanted to make a point about the technicalities of synthetic benchmarks.
