Simulation / Modeling / Design

NVIDIA Deep Learning Inference Platform Performance Study

The NVIDIA deep learning platform spans from the data center to the network’s edge. A new paper describes how the platform delivers giant leaps in performance and efficiency, resulting in dramatic cost savings in the data center and power savings at the edge.

High-level deep learning workflow showing training, then followed by inference.

GPUs have proven to be incredibly effective at solving some of the most complex problems in deep learning, and while the NVIDIA deep learning platform is the standard industry solution for training, its inferencing capability is not as widely understood. Some of the world’s leading enterprises from the data center to the edge have built their inferencing solution on NVIDIA GPUs. Some examples include:

  • Twitter Periscope runs inferencing on GPUs to understand video content in real-time, enabling more sophisticated video searches and user recommendations.
  • Pinterest uses cloud-based GPUs to minimize user wait time (or latency) for its Related Pins service, delivering engaging recommendations based on users’ interests.
  • runs inference-based intelligent video analysis in real time on every frame of video of 1,000 HD video channels, and increased its per-server throughput by 20x.
  • iFLYTEK switched to Tesla GPUs for its Mandarin speech recognition service in China, and is now able to handle 10x the number of concurrent requests, and reduced its operational TCO by 20%.
  • Cisco’s Spark Board and Spark Room Kit, powered by NVIDIA Jetson, are re-inventing the meeting room, enabling wireless 4K video sharing, and using deep learning for voice and facial recognition, as well as enhancing resource planning.
Discuss (0)