Nvidia unveils Titan V with 110 Teraflops of deep learning power
Дек 08 2017 by Joanne Wise
But at almost $3,000 it is really aimed at developers, researchers and companies that wish to create and power AI systems that tap into the parallel processing capabilities of GPU, helping them crunch through massive amounts of data more efficiently than central processing units would.
United States technology firm Nvidia has unveiled its latest processor, the Titan V, that has been named the world's most powerful processor for the PC, using its NvidiaVolta GPU architecture.
And because the Volta is fabricated on a new TSMC 12-nanometer FFN high-performance manufacturing process, which has been customised for the Titan V, it also incorporates a highly tuned 12GB HBM2 memory subsystem for better memory bandwidth utilisation.
Strangely, Nvidia Titan V does not come with NVLink or SLI as it both lacks the support and the actual NVLink connector is blocked with a cooler. It doubles the energy efficiency of the previous generation Pascal design, enabling dramatic boosts in performance in the same power envelope, Nvidia claims. In addition, this graphics card also features 5,120 CUDA Cores, 640 Tensor Cores, and a single-precision floating point compute performance of 15 TFLOPS.
Nvidia founder and CEO Jensen Huang said in a statement -"Our vision for Volta was to push the outer limits of high performance computing and AI".
Nvidia cards are the de facto standard for running machine learning workloads and today, the company added yet another high-end compute-centric card to its line-up: the Titan V. "We broke new ground with its new processor architecture, instructions, numerical formats, memory architecture and processor links", said Huang. The Titan V sports 110 teraflops of raw computing capability, which is nine times that of its predecessor. But the Titan V is way more than just a gaming GPU.
It's a chip that's meant for machine learning researchers, developers, and data scientists who want to be able to build and test machine learning systems on desktop computers.
При любом использовании материалов сайта и дочерних проектов, гиперссылка на обязательна. «» 2007 - 2017 Copyright. Автоматизированное извлечение информации сайта запрещено.