With one week left for the Nvidia GPU Technology conference, it is now conveniently leaked that the upcoming graphics will be named as the GeForce X80 series. The last news about the upcoming flagship Pascal cards was assumed to be named as GTX 1080, followed by GTX 1080 Ti, GTX 1070 and even Titan X’s successor. These chips will be built using the 16nm FinFET process.
The next-generation flagship and high-end graphic cards will be using the GP104 and GP100 chips based on Pascal architecture. It is assumed that the GP104 will use 4,096 CODA cores with 256 TMUs, 128 ROPs and possibly 6GB GDDR5X running via 384 GB/s bandwidth. Its TDP is rated at 175W. The GeForce X80 will be using the GP104 core, and the reference variant would be clocked at 1,000 MHz.
The GP100’s specs are at the higher level as it will have 6,114 CUDA cores, 384 TMUs, 192 ROPs with 8GB GDDR5X. The TDP is expected to be at 225W.
Both the X80 Ti and X80 Titan will be using the GP100. Both cards have a TDP rating of 225 Watts and the reference base clock of 1000MHz, but that’s where the similarity ends. The X80 Ti’s chip will have a reduced spec as one would expect just like with GM200 based GTX 980 Ti and GTX Titan. The X80 Ti’s CUDA cores reduced down to 5120, 320 TMU, 160 ROPs. Another main difference is that while GeForce X80 Ti will be using 6GB GDDR5X, the GeForce will be using the 16GB HBM2. As a result, X80 Ti will have 512 bits memory interface and 512 GB/s memory bandwidth while the X80 Titan will have 4096 bits interface and 1024 GB/s memory Bandwidth. It would be interesting to see the difference in between these GP100, and if the gap will be wider than the previous generation GM200 variants.
Nvidia must have decided to use GDDR5X for consumer-grade graphic cards X80 and X80 Ti because it’s easier to implement even by the AIB manufacturers for their graphic cards. Since previous X80 Titan graphic cards are made by Nvidia, it would probably easy for them to get it fabricate the GP100 with the 16GB HBM2 on-chip die from TSMC.