Originally Posted by Anandtech
The biggest and most visible change is that NVIDIA beefed up the number of various execution units per SM. The 32 CUDA cores from GF100 are now 48 CUDA cores, while the number of SFUs went from 4 to 8 along with the texture units. As a result, per SM GF104 has more compute and more texturing power than a GF100 SM. This is how a “full” GF104 GPU has 384 CUDA cores even though it only has half the number of SMs as GF100
One of the changes of the marchitecture from GF100 to GF104 is defined as stated above in Anandtech's review. Seemingly more theoretical compute and texturing throughput on the GF104 vs GF100.