Among the most prominent changes is the addition of the Advanced Pixel Shader test. This test is supposed to be a Pixel Shader 1.4-compliant test, but according to the posted FAQ, they decided to make the test compliant with versions 1.1 and 1.4. I'll let them explain further:
Q: The Advanced Pixel Shader test is supposed to test 1.4 pixel shader capability. I have a GeForce3 graphics card and can run that test just fine. Does this mean my GF3 can do pixel shader 1.4?I'm not altogether sure if that was a good choice or a bad one. I do know that my GeForce3-equipped system's overall score actually increased with the new version from 7669 to 7959. (You can see my system specs here. I used the 27.20 drivers, and I've overclocked my GeForce 3 to 233/500.) So it would seem that the new test isn't causing scores for GeForce 3 parts to dip.
A: No, your GF3 cannot do pixel shader 1.4! The Advanced Pixel Shader test has a fall-back, using pixel shader 1.0, which is supported by your GF3. Hence the name Advanced Pixel Shader test, and not Pixel Shader 1.4 test. That same water effect can be achieved using pixel shader 1.0 but then the water surface is drawn in two passes, compared to a single pass on pixel shader 1.4 (or higher) hardware.
Q: Will the Advanced Pixel Shader test run faster on graphics cards that support 1.4 pixel shaders?
A: The Advanced Pixel Shader test is what we call a Feature Test, which means that we above all want to present some new technology. It was decided that a fall-back was to be included in addition to the 1.4 pixel shader, since a same looking effect can be achieved using pixel shader 1.1. These two different modes of that same test work a bit differently, and should therefore not be directly compared. Both two modes could be optimized to show more performance either way, but now the test is just optimized for maximum compatibility. Vertex shader performance also affects the score somewhat due to this compatibility optimization.
On the other hand, it seems the GeForce 2 based parts do suffer a score decrease with the new test. My MX200 PCI scores a disreputable 1700 marks, versus ~2200 under the original 3DMark 2001. I imagine the point loss is caused by a variety of factors, but I don't imagine the extra "zero" score in the Advanced Pixel Shader test helped the overall average.
Since the scene is currently devoid of any noteworthy DirectX 8 titles, I question the lower scores 3DMark 2001SE gives to DirectX 7-compliant parts. Providing a benchmark for the future is a cool thing; ignoring the realities of the current market is another.