Nvidia traditionally produces four GPUs every generation, the 0, 4, 6 and 7. The lower this number, the higher the market positioning of the card. GP104 means that we're talking about a GeForce GPU of the Pascal-generation, first generation, and finally the higher mid-range card. With Pascal, Nvidia has added a GPU to this model, namely the GP102. This card is largely equivalent to the P100, but has far fewer double-precision floating point units. This means that the die size is much reduced, while the gaming performance is comparable to what you would get from the P100. The P100 is 610mm², compared to 'just' 471mm² for the GP102. This was a good decision by Nvidia, because it would have been impractical to try to sell such enormous GPUs for prices under 1000 euros - especially considering that TSMC's 16nm process is not very mature yet. A further difference is that the GP102 has the old GDDR5(X) memory controller, whereas the P100 uses HBM. Here you can see the full specifications of the various Pascal GPUs.
|Process||TSMC 16nm||TSMC 16nm||TSMC 16nm||TSMC 16nm||Samsung 14nm|
|Transistors||15,3 billion||12 billion||7,2 billion||4,4 billion||3,3 billion|
|Double precision (FP64) performance||1/2||1/32||1/32||1/32||1/32|
When it comes to the VRM's, the PEG-connectors and the TDP, there is no difference between the two varieties of the GP102. Both the Titan X (Pascal) and that GTX 1080 Ti Founders Edition have a PEG6- and a PEG8-connector, making for an official maximum power draw of 300 watt. The TDP of 250 watt is also identical.
This is not the case for the video-outputs. While the Titan X (Pascal) had three Displayport 1.4 and one HDMI 2.0 outputs, along with an old-fashioned DVI-port, the DVI has been eliminated in the GTX 1080 Ti. Instead the card has more openings for blowing out the hot air, which might improve the cooling.