NVIDIA DGX-1 with Tesla V100 system architecture

"The NVIDIA® DGX-1TM with Tesla V100 ( Figure 1) is an integrated system for deep learning. DGX-1 features 8 NVIDIA® Tesla® V100 GPU accelerators connect through NVIDIA® NVLinkTM, the NVIDIA highperformance GPU interconnect, in a hybrid cube-mesh network. Together with dual-socket Intel Xeon CPUs and four 100 Gb InfiniBand network interface cards, DGX-1 provides unprecedented performance for deep learning training. Moreover, the DGX-1 system software, powerful libraries, and NVLink network are tuned for scaling up deep learning across all eight Tesla V100 GPUs to provide a flexible, maximum performance platform for the development and deployment of deep learning applications in both production and research settings.

DGX-1 with V100 GPUs achieves dramatically higher throughput than DGX-1 with previous-generation NVIDIA Tesla P100 GPUs, achieving up to 3.1x faster deep learning training for convolutional neural networks. High-performance NVLink GPU interconnect improves scalability of deep learning training, improving recurrent neural network training performance by up to 1.5x compared to slower PCIe interconnect. More productivity and performance benefits come from the fact that DGX-1 is an integrated system, with a complete optimized software platform aimed at deep learning that ensures DGX-1 outperforms similar off-the-shelf systems."