In most current games, GeForce2 and GeForce2MX cards are bottlenecked by memory bandwidth at 1024x768x32. The GeForce2MX/MX400 is especially limited by its 128-bit SDR SDRAM compared to the 128-bit DDR memory on the GeForce2 GTS/Pro/Ti. The GeForce2MX200 is doubly-castrated by its puny 64-bit SDR memory bus.
When you enable hardware T&L, even more data is sent back and forth for each pixel. Your card is probably already bottlenecked by its available bandwidth in this demo, so you end up going slower.