As you can see from the specifications above, the GTX780 has an increase of SMXs over the GTX680 from 8 to 12, 2 less that of the GTX TITANs 14 SMXs (1 less than the 'full fat' 15 (5x3) available on GK110). We also see a 50% increase in CUDA cores over the GTX680 to 2304 which sees the texture unit total rise to 192. Raster Operations are also enhanced by a third to 48, again matching the GTX TITAN. Interestingly the GTX780 is clocked slightly higher than the GTX TITAN but this isn't quite enough to enable the GTX780 to match the TITANs 187.5 GT/s count although 165.7 GT/s is well above the GTX680's 128.8.

The GTX780 also 'makes do' with half the memory of the GTX TITAN but still 1GB more than the GTX680. The good news however is that the GTX780 has inherited the 384-bit memory interface (6x64bit) which gives the GTX780 a healthy memory bandwidth of 288.4 GB/s.

The architecture of the NVIDIA GTX GTX780 core; the GK110, is an evolution of the GTX680's GK104 Kepler core but is trimmed back from the GTX TITAN's GK110, losing 2xSMX. So doing the math 2688-384 = 2304 Cuda Cores. All cores are based on Kepler architecture and all are built upon the 28nm ASIC. However the GK110 features 7.1 billion transistors compared to the 3.54 billion found in the GTX680 and the 4.4 billion of the AMD HD7970.



Above we can see a breakdown of an individual SMX. We were happy to see NVIDIA increasing the memory size to 3GB of GDDR5 along with the 384-bit memory controller. This gives a healthy boost over the 2GB found on the GTX680. 3GB is plenty for todays gaming however with games becomming increasingly demanding, we would have liked to have seen 4GB of GDDR5 which would have positioned the GTX780 neatly between the outgoing and flagship models.