I believe this tweak is helpfull on NVidia system with a low amount of video memory but not really helpfull on 256meg cards.
Reasoning:
A pixel on the primary display surface takes up 4 bytes of memory. In addition, each backbuffer also uses 4 bytes per pixel. With tripple buffering enabled, thats 4 + 4*3 = 16 bytes per pixel. With single buffering, it will only be 4 + 4 = 8 bytes per pixel. Saving 50% of the overhead for pixel color information.
On my system specifically with its 1280x1024 native resolution (LCD) its a savings of 10 megs of video memory. 10 megs isnt much on a 256 meg card like I have, but is quite a lot on older 64meg cards.
The saved memory can then be used to store textures or geometry data that would otherwise have to be shuttled over the PCI/AGP bus every frame. This not only makes the video card more efficient, it can also make other periferals (and even system memory) more responsive because there is less front side bus traffic.