NVIDIA Tesla V100

From scientific discovery to artificial intelligence, high performance computing (HPC) is an important pillar that fuels the progress of humanity. Modern HPC data centers are currently solving some of the greatest challenges facing the world today. With traditional CPUs no longer delivering the performance gains they used to, the path forward for HPC data centers is GPUaccelerated computing.

NVIDIA® Tesla® is the leading platform for accelerated computing and powers some of the largest data centers in the world—delivering significantly higher throughput while saving money.

NVIDIA Tesla V100 powered by NVIDIA Volta™ architecture is the computational engine for scientific computing and artificial intelligence. Here are three powerful reasons to deploy NVIDIA Tesla V100 GPUs to your data center.

The AI revolution is here, and every data center should be equipped for it. AI is the engine behind consumer services we use every day, like web searches and video recommendations. AI extends traditional HPC by allowing researchers to analyze large volumes of data for rapid insights where simulation alone cannot fully predict the real world, in domains like bioinformatics, drug discovery, and highenergy physics.

AI IS HELPING RESEARCHERS CREATE CLEAN ENERGY USING FUSION REACTORS

Princeton University is using AI to predict and respond to disruptions in a tokamak reactor (a magnetic fusion reactor device).

NVIDIA Tesla V100 is the computational engine driving the AI revolution and enabling HPC breakthroughs. For example, researchers at the University of Florida University and University of North Carolina leveraged GPU deep learning to develop ANAKIN-ME (ANI) to reproduce molecular energy surfaces at extremely high (DFT) accuracy and 1-10/millionths of the cost of current computational methods.

AI SPEEDS DRUG DEVELOPMENT

Today’s CPU-based methods can process as many as two million candidates per year, but researchers at the Universities of Florida and North Carolina have developed a way to use GPU-based deep learning to accomplish the same work in under two minutes.

Over 550 HPC applications are already GPUoptimized in a wide range of areas including quantum chemistry, molecular dynamics, climate and weather, and more.

In fact, an independent study by Intersect360 Research shows that 70% of the most popular HPC applications, including 10 of the top 10 have built-in support for GPUs.

With the most popular HPC applications and all deep learning frameworks GPUaccelerated, every HPC customer would see most of their data center workload benefit from GPU-accelerated computing.

Data center managers all face the same challenge: how to meet the demand for computing resources that often exceed available cycles in the system.

NVIDIA Tesla V100 dramatically boosts the throughput of your data center with fewer nodes, completing more jobs and improving data center efficiency.

A single server node with V100 GPUs can replace over 60 CPU nodes. For example, for SPECFEM3D, a single node with four V100’s will do the work of 53 dual-socket CPU nodes while for NAMD a single V100 node can replace 13 CPU nodes. With lower networking, power, and rack space overheads, accelerated nodes provide higher application throughput at substantially reduced costs.

Volta Architecture

By pairing CUDA Cores and Tensor Cores within a unified architecture, a single server with Tesla V100 GPUs can replace hundreds of commodity CPU servers for traditional HPC and Deep Learning.

Tensore Core

Equipped with 640 Tensor Cores, Tesla V100 delivers 125 TeraFLOPS of deep learning performance. That’s 12X Tensor FLOPS for DL Training, and 6X Tensor FLOPS for DL Inference when compared to NVIDIA Pascal™ GPUs.

Next Generation NVLINK

NVIDIA NVLink in Tesla V100 delivers 2X higher throughput compared to the previous generation. Up to eight Tesla V100 accelerators can be interconnected at up to 300GB/s to unleash the highest application performance possible on a single server.

Maximum Efficiency

The new maximum efficiency mode allows data centers to achieve up to 40% higher compute capacity per rack within the existing power budget. In this mode, Tesla V100 runs at peak processing efficiency, providing up to 80% of the performance at half the power consumption.

HBM2

With a combination of improved raw bandwidth of 900GB/s and higher DRAM utilization efficiency at 95%, Tesla V100 delivers 1.5X higher memory bandwidth over Pascal GPUs as measured on STREAM. Tesla V100 is now available in a 32GB configuration that doubles the memory of the standard 16GB offering.

Programmability

Tesla V100 is architected from the ground up to simplify programmability. Its new independent thread scheduling enables finer-grain synchronization and improves GPU utilization by sharing resources among small jobs.