boredgunner

Fermi is what they call the 400 series cards.  The SE is the newer model with 1GB VRAM, 256 bit bus width and 288 CUDA cores.  Which one is better?  Depends on the use for the card.  The extra memory bandwidth of the SE will help at larger resolutions while the 768 MB version has a higher pixel fill rate and texture fill rate.  It will perform better at lower resolutions and it is better for any GPGPU application such as CUDA applications (Folding @ Home, PhysX) and DirectCompute applications.

boredgunner

The GTX 460 768 MB would most likely be a better choice, unless you plan on upgrading your monitor.  Alternatively there is the GTX 460 1GB which has the best of both worlds.  Same memory configuration as the GTX 460 SE but same shader count and clock speeds as the GTX 460 768 MB (and 8 more ROPs).  But at that price you're close to the AMD 6850 which is an even better performer with greater efficiency.

boredgunner

Between what?  The two you posted?  Look at the memory configuration and resources.  The GTX 460 768 MB has 336 CUDA cores/56 TMUs/24 ROPs while the SE only has 288 CUDA cores/48 TMUs/32 ROPs but a better memory configuration.  The GTX 460 1GB has 336 CUDA cores/56 TMUs/32 ROPs and the same memory configuration as the SE, plus a small clock speed advantage.  You can see the full specs for those cards, and much more here.

http://www.techarp.com/article/Desktop_GPU_Comparison/nvidia_4_big.png

relance

Tiger direct has their info wrong, the amount of Stream Processors for the 1GB SE should be 288, not 336. 
http://www.evga.com/products/moreinfo.asp?pn=01G-P3-1366-TR
http://www.evga.com/products/moreInfo.asp?pn=768-P3-1360-TR
Those links have the right specs for each card. I too was thinking about getting the 1GB SE, and returning my 768MB, but I decided to keep the 768MB as I most likely wont be playing at resolutions of 1680x1050 or higher, so the extra processors that the 768MB has over the 1GB SE would benefit me more.