NVIDIA Introduces Titan V For Machine Learning Acceleration On The PC

NVIDIA

NVIDIA TITAN V

NVIDIA created the “TITAN” line of products years ago to represent “the best” of everything the company brings to the PC platform. If you want the best and can afford it, NVIDIA suggests TITAN. TITAN’s target market has shifted and will continue to evolve.  In the past TITAN targeted gamers (TITAN X) or machine learning scientists (TITAN XP). Today, NVIDIA dropped a bombshell at the annual NIPS (Neural Information Processing Systems) conference in Long Beach, CA, with the introduction of its new “TITAN V” PC GPU. I guess we shouldn’t be surprised at this point—NVIDIA has been firing on all cylinders the past few years and has been surprising just about everyone. Let’s take a deeper look at the TITAN V – who it is targeted at and what it has going for it.

Targeting machine learning scientists who use desktop PCs

TITAN V is targeted at machine learning scientists who want to conveniently buy the card and install it into their desktop PC. This means the researcher doesn’t need a special server, storage or networking. Machine learning workloads favor heavy-duty matrix math operations which require massive memory bandwidth and this is what TITAN V delivers.

Driven by Volta architecture

Driven by NVIDIA’s latest GPU architecture, Volta, NVIDIA says the TITAN V’s 21.1 billion transistors are capable of delivering 110 teraflops of performance (for reference, that’s 9x times the deep learning computing horsepower of its predecessor). This immense amount of performance makes the TITAN V ideally suited users looking to explore computational processing for scientific simulation and other deep learning/AI applications on their desktop PCs. This is different from Tesla that is targeted at server systems and deep learning appliances like the DGX Station.

In addition to the muscle, the Titan V is also highly energy efficient. Its Volta architecture underwent a serious revamping of the SM processor design at the core of the GPU, which should providetwice the  energy of efficiencyPascal. This allows for major boosts in FP32 and FP64 performance, within the same power envelope.

The Titan V also sports the same new Tensor Cores that are in Tesla GPUs which are geared specifically towards deep learning purposes. NVIDIA also says the Volta architecture is more efficient at HPC, thanks to its independent parallel integer and floating-point data paths.

Powered by WPeMatico