Provides 110 teraflops of processing power, an improvement of approximately 9 times compared to the previous product.

NVIDIA unveiled the Titan V, a powerful PC GPU based on the cutting-edge NVIDIA Volta GPU architecture, at a conference affiliated with the Neural Information Processing Systems (NIPS).

NVIDIA founder and CEO Jensen Huang personally announced the Titan V, a GPU specialized for computational processing in scientific simulations, at the NIPS conference. With 21.1 billion transistors, the Titan V delivers 110 teraflops of processing power, an improvement of approximately nine times over its predecessor, while boasting outstanding energy efficiency.

CEO Jensen Huang stated, “With Volta, we envisioned breaking through the limits of High Performance Computing (HPC) and AI, pioneering new frontiers through new processor architectures, processing methods, number formats, memory architectures, and processor links,” adding, “NVIDIA will provide Volta-based Titan V to researchers and scientists around the world, and we look forward to the groundbreaking discoveries they will make.”

NVIDIA's Supercomputing GPU Architecture Now Enters the PC Realm

The most significant feature of the Titan V Volta architecture is the extensive redesign of the streaming multiprocessor at the center of the GPU. The energy efficiency of the previous generation Pascal design has also more than doubled, dramatically improving performance while maintaining the same power consumption.

The new Tensor Cores, designed for deep learning, deliver up to nine times higher teraflop performance. Additionally, Volta features independent parallel integer and floating-point data paths, demonstrating enhanced efficiency in mixed-operation workloads and computation processing. Performance has been significantly improved and programming simplified through the new integrated L1 data cache and shared memory units. Manufactured using a new TSMC 12-nanometer FFN high-performance process customized for NVIDIA, the Titan V is equipped with Volta's 12GB HBM2 memory subsystem, which has been high-performance tuned to utilize advanced memory bandwidth.

Titan V users can immediately access advanced AI, deep learning, and HPC software optimized for GPUs after signing up for a free NVIDIA GPU Cloud account. The free container registry includes NVIDIA-optimized deep learning frameworks, third-party managed HPC applications, NVIDIA HPC visualization tools, and NVIDIA TensorRT™ inference optimization features.