NVIDIA decided to use the same name for its newer processor, except equip it with much faster GPU, Pascal GP102. This processor will features 3584 CUDA cores, 224 TMUs and 96 ROPs. The card is equipped with dazzeling 12 GB of GDDR5X memory across 384-bit interface.
NVIDIA has already revealed the clock speeds of this model, which are 1417 MHz for base and 1531 MHz for boost clock. The memory is clocked at the same speed as GTX 1080, 10 GHz, but thanks to wider bus we have 480 GB/s of bandwidth.
This clock speed and such number of unified cores gives almost maximum 11 TFLOPs of raw computing power in 32-bit operations. It means that new TITAN X is up to around 60% faster than it’s predecessor.
The card is targeting 250W TDP, while being equipped with two power connectors (8+6-pin) unlike the 1070 and 1080, so just as many other flagship cards from NVIDIA for the past 2 years.
TITAN X will supports SLI technology through HB SLI bridges, but it’s unknown how many cards can be connected together.