GTC 2017: NVIDIA GPU Cloud Platform (NVIDIA keynote part 10)

TL;DR

Volta revolutionizes deep learning with Tensor Core for training and inferencing, enhanced Graph Analytics, and containerized GPU cloud computing.

Transcript

let me talk about inferencing so now we've created this network and it's taken hours and hours of deep learning training on DG x1 or in in the Amazon Cloud or in the azure cloud with all these GPUs now that you have this network this network can is now ready to be deployed and that network is still very computationally intensive and we need to figu... Read More

Key Insights

• 🧑‍🏭 Volta's Tensor Core enhances training and inferencing speed by a factor of 6 and 12, respectively.
• 😕 Graph Analytics in Volta optimizes neural networks by fusing and sharing mathematical operations efficiently.
• 😶‍🌫️ Containerized GPU cloud platform simplifies deep learning deployment and management, offering optimized frameworks for accelerated computing.
• 😶‍🌫️ NVIDIA GPU cloud provides a user-friendly interface for running deep learning jobs in cloud, supporting various frameworks and optimized containers.
• 😘 Volta's inferencing performance surpasses Broadwell and Skylake CPUs significantly, delivering high throughput and low latency.
• 🎨 Volta's efficiency is highlighted by the compact design of the Tesla V100 accelerator, offering substantial savings and increased data center throughput.
• 😶‍🌫️ The NVIDIA GPU cloud platform supports accelerated computing environments, such as cloud and DGX supercomputers, enhancing scalability and performance.

Questions & Answers

Q: How does Volta improve inferencing speed?

Volta accelerates inferencing by integrating Tensor Core, increasing throughput by a factor of 6, and utilizing tensor runtime optimization for efficient processing.

Q: What role does Graph Analytics play in optimizing neural networks?

Graph Analytics in Volta identifies opportunities to fuse and share mathematical operations within neural networks, enhancing efficiency and performance.

Q: How does the containerized GPU cloud platform simplify deep learning deployment?

The NVIDIA GPU cloud offers a registry of optimized frameworks for easy deployment, allowing users to access scalable GPU resources for accelerated deep learning tasks.

Summary & Key Takeaways

• Volta, powered by Tensor Core, boosts training and inferencing speed significantly.

• Graph Analytics optimizes neural networks by fusing and sharing mathematical operations.

• Containerized GPU cloud platform simplifies deep learning deployment and management.