NVIDIA Tesla V100S

Description

Maximum efficiency mode

The maximum efficiency mode allows data centers to achieve high compute capacity per rack within the existing power budget. In this mode, Tesla V100 runs at peak processing efficiency, providing advanced performance at half the power consumption.

Tensor Core

Equipped with 640 Tensor Cores, Tesla V100 delivers 120 TeraFLOPS of deep learning performance.

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.

Next generation NVLink

NVIDIA NVLink in Tesla V100 delivers high throughput - Tesla V100 accelerators can be interconnected to unleash better application performance on a single server.

High memory bandwidth

With a combination of improved raw bandwidth and high DRAM utilization efficiency Tesla V100 delivers high memory bandwidth over Pascal GPUs.

Programmability

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