Implementing deep learning at the level of 100 CPUs
One Tesla V100 GPU replaces hundreds of CPUs

NVIDIA announced Volta™, the most powerful GPU computing architecture to lead the advancement of artificial intelligence and high-performance computing, during the keynote speech at the NVIDIA GPU Technology Conference (GTC).

During this keynote, the NVIDIA® Tesla® V100 Data Center GPU, the first Volta-based processor, was also announced. The Tesla V100 Data Center GPU supports the exceptional speed and scalability required for artificial intelligence inference and training, and accelerates high-performance computing and graphics workloads.

NVIDIA founder and CEO Jensen Huang stated, “Artificial intelligence will automate intelligence and bring about new social progress unprecedented since the Industrial Revolution.” He also said, “Deep learning, an innovative AI approach to building learnable computer software, requires infinite processing power. To meet this demand, thousands of NVIDIA engineers have dedicated themselves to developing Volta over the past three years.”

NVIDIA's 7th generation GPU architecture, Volta, is built with 21 billion transistors and enables deep learning with performance equivalent to 100 CPUs.

As networks become more advanced, data centers must provide exponentially improved processing power. Furthermore, they must possess efficient scalability to facilitate the rapid introduction of highly accurate AI-based services, such as natural language virtual assistants (VA) and personalized search and recommendation systems.

Volta is expected to set a new standard for high-performance computing by providing an outstanding platform for computational science and data science for new insights into high-performance computing systems. By combining CUDA® Cores and the new Volta Tensor Cores into a unified architecture, a single server equipped with a Tesla V100 GPU can replace hundreds of general-purpose CPUs required for existing high-performance computing.