DATA CENTER
ACCELERATION
Dave Salvator, Senior Manager, Product Management, NVIDIA
EVOLUTION OF COMPUTING
AI & IOT
Deep Learning, GPU
100s of billions of devices
Mobile-Cloud
iPhone, Amazon AWS
2.5 billion mobile users
PC Internet
WinTel, Yahoo!
1 billion PC users
1995 2005 2015
22
NVIDIA
“THE AI COMPUTING COMPANY”
GPU Computing Computer Graphics Artificial Intelligence
3
RISE OF NVIDIA GPU COMPUTING
109
APPLICATIONS 108 1000X
GPU-Computing perf
In 10
107 2X per year
ALGORITHMS years
106
SYSTEMS 1.1X per year
105
CUDA 104
103
1.5X per year
ARCHITECTURE 102
Single-threaded perf
1980 1990 2000 2010 2020
Original data up to the year 2010 collected and plotted by M. Horowitz, F. Labonte, O. Shacham, 4
K. Olukotun, L. Hammond, and C. Batten New plot and data collected for 2010-2015 by K. Rupp
BEYOND MOORE’S LAW
Progress Of Stack In 7 Years
2013 2020
cuBLAS: 5.0 cuBLAS: 11.0
cuFFT: 5.0 cuFFT: 11.0
cuRAND: 5.0 cuRAND: 11.0
cuSOLVER: 11.0
Relative Performance
cuSPARSE: 5.0
NPP: 5.0 GPU-Accelerated cuSPARSE: 11.0
Computing
Thrust: 1.5.3 NPP: 1`1.0
CUDA: 5.0 Thrust: 1.9.0
Resource Mgr: r304 CUDA: 11.0
Moore’s Law
Base OS: CentOS 6.2 Resource Mgr: r384
Base OS: Ubuntu 16.04
CPU
2013 2014 2015 2016 2017 2018 2019 2020
Accelerated Server Accelerated Server
With Fermi with Ampere
5
25 YEARS OF ACCELERATED COMPUTING
DEVELOPERS ++
DEVELOPMENT
GPU
INSTALLED PERFORMANCE ++
ACCELERATION BASE ++
DPU
COMPUTE
CUDA
EVERYWHERE
CPU NETWORKING
X-factor Speed-up Full Stack Data Center Scale One Architecture
6
NVIDIA DATACENTER PLATFORM
BUSINESS Customer Patient Fraud Quality Industrial Precision Molecular
++
APPLICATIONS Engagement Diagnostics Detection Assurance Automation Marketing Simulations
NGC
SMART CITY CONVERSATIONAL AI AUTONOMOUS RECOMMENDATION HEALTHCARE ++ OPERATIONS
APPLICATION VEHICLES SYSTEMS
SOFTWARE HUB FRAMEWORKS Merlin ...
Metropolis Jarvis Drive Clara
VIRTUAL GPU SW
TRITON
INFERENCE
ML & DATA ANALYTICS AI TRAINING & INFERENCE HIGH PERFORMANCE RENDERING & SERVER
Certified FLEET
Con
Containers
DEVELOPER COMPUTING VISUALIZATION COMMAND
TensorRT
TOOLKITS IndeX OptiX
NVIDIA GPU
NVIDIA HPC SDK MDL
CloudXR Operator
Pre-trained Models
COMPUTE MANAGEMENT
SDKs ACCELERATION NETWORKING, STORAGE & SECURITY
LIBRARIES CUDA-X DOCA MAGNUM IO
NVIDIA CERTIFIED MONITORING
SERVERS & DGX
EGX
VALIDATED HGX
CLOUD DCGM
SOLUTIONS
CSP Instances
Purpose Built Mainstream & Edge
HARDWARE
UFM
TECHNOLOGIES
GPU NVSwitch BF DPU SMART NIC NVIDIA Switch
7
AMAZING EXPANSION OF NVIDIA ECOSYSTEM
Apps for Every Industry Reaching Billions of Users
80 2.3M
New SDKs Developers
6M
DLSS 2.1 CUDA Downloads in 2020
RTX DI
OptiX 7.2
AERIAL
HPC SDK 20.9 RTX HPC RAPIDS AI CLARA METRO DRIVE ISAAC 5G
RAPIDS 0.16
Parabricks 3.5
DeepStream 5.0 1,800
GPU-Accelerated Applications
NSIGHT 2020.5
cuDNN 8.03 CUDA-X
TensorRT 7.2
CUDA 11.1 CUDA
6,500
NCCL 2.7.8 AI Startups
MAGNUM IO
GPUDirect Storage
10
11
12
13
14
15
16
17
18
20 19
T
ES
20
20
20
20
20
20
20
20
20
20
20
COMPLETE SOFTWARE STACK GROWING ECOSYSTEM
