Seemingly out of nowhere NVIDIA has announced its new flagship graphics card, the TITAN V. The card makes use of the “Volta” GV100 graphics processor, the same one that drives NVIDIA’s Tesla V100 HPC accelerator. The card has a multi-chip module design with the GPU die and three HBM2 memory stacks sharing the same package.
The GPU die has been built on the 12nm FinFET+ process by TSMC. The TITAN V maxes out the GV100 silicon, featuring 5120 CUDA cores, 640 Tensor cores (specialized units that accelerate neural-net building/training), and 320 TMUs. The CUDA cores are spread across 80 streaming multiprocessors (64 CUDA cores per SM), spread across 6 graphics processing clusters (GPCs). There is 12GB of HBM2 across a 3072-bit wide memory interface.
The GPU core is clocked at 1200 MHz, with a GPU Boost up to 1455 MHz, with the HBM2 memory running at 850 MHz. That equates to 652.8 GB/s of memory bandwidth (1.70 Gbps stacks). The card will draw power from a combination of 6-pin and 8-pin PCIe power connectors. As far as display outputs go you have three DisplayPort, and one HDMI.
The card will come in with a price tag of $2999, and is exclusively available from NVIDIA’s online store.