To call these results surprising would be an understatement, and a more detailed examination of 3DMark 2001 is definitely required. Thankfully, 3DMark 2001 offers a detailed list of the individual test scores through their result viewer. What we found was that in many basic performance indicators the GeForce2 Ultra attained the higher ratings. Since the GeForce3 shares the same basic architecture, and the GeForce2 Ultra is clocked higher, this was to be expected. Moving farther down the list we also found that the only card supporting the 3DMark 2001 Pixel Shader was the GeForce3. The stated fill-rate (multi-texture) numbers were also a bit strange and well in excess of the GeForce3's physical specifications. Perhaps 3DMark 2001 is taking into account the GeForce3's quad pixel shader when calculating fill rate?

For a detailed look at the database of the results, check the screenshot below:

The other major discrepancy is that the GeForce3 is the sole video card that can run the Nature 4 test. The results of this test are factored into the final 3DMark 2001 rating, so this gives the GeForce3 a distinct advantage over the competition. While the Nature 4 portion of 3DMark 2001 does gives extra points for utilizing hardware pixel shaders, it is somewhat troubling that one specific feature would affect an entire segment of the benchmark. Instead of defaulting to DirectX 8 software mode, the test simply refuses to run on any card but the GeForce3.