2018年英伟达投资者日大会PPT-2018.3.27-76页-21mb

NVIDIA Technology Summary

Core Content

NVIDIA is a leading technology company that focuses on graphics, science, AI, and robotics. The company has introduced several groundbreaking technologies and platforms to enhance real-time rendering, AI performance, and autonomous systems.

Key Technologies

Graphics & Rendering

• Screen-Space Ambient Occlusion: A technique for realistic lighting effects in real-time rendering.
• Baked Lighting: Precomputed lighting for efficient rendering.
• Global Illumination: Advanced lighting simulation for realistic visuals.
• Screen-Space Reflections: Real-time reflection rendering.
• Environment Maps: Used for simulating reflections and lighting.
• Ray Traced Reflections: High-fidelity reflection rendering using ray tracing.
• Screen-Space Refraction: Real-time refraction simulation.
• Depth Sorting: Ensures correct rendering order of objects.
• Caustics: Simulates light focusing effects.
• Subsurface Shading Approximation: Simulates light interaction with translucent materials.
• Subsurface Scattering: Simulates light diffusion in materials like skin or wax.

AI & Deep Learning

• NVIDIA AI Platform: Offers a comprehensive ecosystem for AI development and deployment, including tools like TensorRT, CUDA, cuDNN, and DGX systems.
• NVIDIA AI Inference: Provides optimized AI inference solutions for various applications such as video analytics, speech processing, recommendation systems, mapping, automotive, robotics, smart cities, etail, healthcare, and manufacturing.
• NVIDIA GPU Cloud: Offers GPU-optimized containers for cloud computing, supporting platforms like AWS, GCP, Azure, Oracle Cloud, and AliCloud.
• NVIDIA DGX-2: The world's largest GPU, featuring 16 Tesla V100 32GB GPUs connected via NVSwitch, with 512GB HBM2 memory and 2,000 TFLOPS Tensor Cores.
• NVIDIA DRIVE: An end-to-end platform for autonomous vehicles, including DRIVE PX 2, DRIVE Xavier, DRIVE PX Parker, and DRIVE Pegasus.
• NVIDIA DRIVE Sim and Constellation: A virtual reality simulation system for AV validation, capable of simulating rare and difficult conditions and running regression tests.
• NVIDIA Clara - Medical Imaging Supercomputer: A platform for medical imaging and visualization, using DL-based image reconstruction and brain segmentation.

Key Products

NVIDIA RTX Technology

• RTX Technology: Enables interactive ray tracing and real-time rendering.
• NVIDIA Quadro GV100: A high-performance GPU with 64GB HBM2 memory, 10,240 CUDA cores, and 236 TFLOPS Tensor Cores.
• NVIDIA Tesla V100: A powerful GPU with 32GB memory, used in HPC and deep learning applications.
• NVIDIA TITAN V: A high-end GPU for deep learning and AI research.

NVIDIA DGX Systems

• DGX-1 and DGX Station: High-performance systems for deep learning and HPC.
• DGX-2: Offers 2 PFLOPS performance, 10 kW power, and 1/8 the cost of traditional systems.

Key Partnerships & Collaborations

NVIDIA collaborates with various healthcare providers, startups, and imaging companies, including:

• Arterys
• Brigham and Women's Hospital
• Mayo Clinic
• Memorial Sloan Kettering Cancer Center
• Nuance of Medicine
• UCL
• Universiteit Antwerpen
• University of Basel
• GE Healthcare
• PerkinElmer
• Siemens Healthineers

Research & Innovation

NVIDIA has a large research team (200 researchers) working on:

• Graphics
• Deep Learning
• Robotics
• Computer Vision
• Parallel Architectures
• Programming Systems
• Circuits & VLSI
• Networks

Performance & Efficiency

• NVIDIA RTX Technology allows for interactive ray tracing and real-time rendering.
• NVIDIA DGX-2 provides big savings in cost, space, and power compared to traditional systems.
• NVIDIA GPU Cloud supports GPU-optimized containers for cloud computing, with 30 containers and 20,000+ registered organizations.
• NVIDIA DRIVE Sim and Constellation enables billions of virtual miles of testing, with 10,000 constellations driving 3 billion miles annually.

Quotes

• "With RTX we can now do ray tracing renders interactively. It's just fantastic!" – Sebastien Guichou, CTO, Isotropix
• "NVIDIA RTX opens the door to make real-time ray tracing a reality!" – Kim Libreri, CTO, Epic Games

Roadmap & Future Goals

• NVIDIA DRIVE Roadmap: Includes Orin and next-generation systems for autonomous driving.
• NVIDIA AI Platform: Continues to evolve with new models and technologies, such as WaveNet, LSTM, GRU, and Capsule Nets.
• NVIDIA GPU Cloud: Expands its cloud offerings with support for Kubernetes, TensorRT, TensorFlow, and ONNX.

Summary

NVIDIA continues to push the boundaries of graphics, AI, and autonomous systems through its advanced GPU technology, deep learning platforms, and supercomputing solutions. The company's innovations, such as RTX, DGX-2, and DRIVE, are transforming industries like gaming, media & entertainment, product design, architecture, healthcare, and automotive. With a strong research foundation and extensive partnerships, NVIDIA is well-positioned to lead in the future of computing and AI-driven technologies.