NVIDIA | Investor Day Presentation Deck | 97 slides

NVIDIA · 2020-05
arc beats above · slides in the middle · loops below · scroll → 2 LOOPS
SETUP TENSION ANALYSIS EVIDENCE RESOLUTION APPENDIX
HOVER FOR DETAILS · CLICK A SLIDE FOR FULLSCREEN · STEP 5
No image bound to this slide. Wire up imgproxy to display the rendered slide JPEG.

Deck intelligence map

5
coverage by narrative range · generated from this deck JSON

Slide inventory

97
every slide · same image gating as the playbook
01
Slide 1
front_matter
Open slide detailBeat · Setup
02
The slide uses a central text overlay on a collage of diverse frontline workers.summarize
Open slide detailBeat · Setup
03
The slide uses a horizontal timeline-like structure at the bottom to categorize the applications into Containment, Mitigation, Treatment, and Tracking & Monitorillustrate_case
Open slide detailBeat · Setup
04
The slide uses a 3D product display layout to associate specific hardware platforms with various industry use cases (robotics, autonomous driving, etc.).establish_context
Open slide detailBeat · Setup
05
This appears to be a transition or title slide background with no text content.filler
Open slide detailBeat · Setup
06
The 'One Architecture' section depicts a flywheel effect.summarize
Open slide detailBeat · Setup
07
The rightmost diagram represents a network effects flywheel.present_framework
08
The slide uses a black background with reflections of the logos at the bottom.transition
09
The slide combines a stack diagram of software layers with a bar chart showing developer growth over time.summarize
10
The slide features a photo of Jensen Huang holding an RTX card on stage.cite_precedent
11
The diagram shows a pipeline: Ray Tracing -> Deep Learning (Neural Network) -> Upscaled Output.present_framework
12
The diagram shows a neural network training process involving a supercomputer.present_solution
13
The diagram shows a neural network processing ray-traced inputs to produce a final image, facilitated by NVIDIA drivers and NGX technology.present_solution
14
Contains Unreal Engine logo and NVIDIA branding.filler
15
The slide features the Unreal Engine logo in the bottom left and a green callout box in the bottom right indicating 'Native 720p'.filler
16
The slide features the Unreal Engine logo and a green NVIDIA DLSS 1.0 callout box.other
17
The slide features the Unreal Engine logo and an NVIDIA DLSS 2.0 callout box.other
18
Includes Unreal Engine logo and NVIDIA-style 'Native 1080p' badge.filler
19
Contains Unreal Engine logo and NVIDIA-style green branding tag.filler
20
The slide contains the Unreal Engine logo in the bottom left and a 'Native 540p' resolution badge in the bottom right.filler
21
The slide features the Unreal Engine logo and a green NVIDIA DLSS 2.0 callout box indicating an upscaling resolution from 540p to 1080p.other
22
The slide uses a background image of a Minecraft scene to demonstrate the visual quality alongside the performance metric.present_solution
Open slide detailBeat · Before-After
23
The slide uses a dark-to-light gradient to emphasize the visual quality of the game screenshot on the right.illustrate_case
Open slide detailBeat · Before-After
24
This image appears to be a decorative or atmospheric slide, likely used as a transition or background.filler
Open slide detailBeat · Before-After
25
The slide uses a visual metaphor of a globe connected to various studios and a cluster of software icons to represent the fragmented and complex nature of the 3frame_problem
Open slide detailBeat · Problem Statement
26
The slide uses a hub-and-spoke visual metaphor to show software integration into the Omniverse platform.present_framework
Open slide detailBeat · Problem Statement
27
The slide illustrates the NVIDIA Omniverse ecosystem architecture, connecting hardware (server) to software (VMs) to end-user applications and global visualizatpresent_solution
Open slide detailBeat · Problem Statement
28
This slide contains no text or data; it is purely atmospheric imagery.filler
Open slide detailBeat · Problem Statement
29
Includes logos of partner hardware vendors (BOXX, Dell, HPE, Supermicro) and a visual representation of a project by Zaha Hadid Architects.front_matter
Open slide detailBeat · Propose Solution
30
The slide uses a modular layout to showcase different pillars of the NVIDIA HPC strategy.analyze_data
Open slide detailBeat · Propose Solution
31
The slide maps specific software stacks (Spark, TensorFlow, PyTorch, cuDNN, Magnum IO, TensorRT, Triton) to the ML pipeline stages.present_framework
Open slide detailBeat · Propose Solution
32
The slide contrasts historical data growth with a modern ML stack (Spark, TensorFlow/PyTorch, cuDNN, TensorRT).present_framework
Open slide detailBeat · Propose Solution
33
The slide uses a circular inset to show the software stack layers and a process flow diagram to show the integration with Spark 3.0.present_solution
Open slide detailBeat · Propose Solution
34
The slide highlights performance metrics (17 GB/s, $1M cost) alongside a high-level data processing architecture.present_framework
Open slide detailBeat · Propose Solution
35
The slide uses a process flow diagram to illustrate the integration of hardware and software stacks.present_framework
Open slide detailBeat · Propose Solution
36
The slide highlights a 5X price-performance improvement and specific throughput metrics for a TPCx-BB benchmark.present_solution
Open slide detailBeat · Propose Solution
37
The slide uses a value-chain style process flow to illustrate software stack optimization.present_framework
Open slide detailBeat · Propose Solution
38
The slide highlights efficiency gains (1/5th cost, 1/3rd power) and performance metrics (163 GB/s throughput) for a specific TPCx-BB benchmark.present_solution
Open slide detailBeat · Propose Solution
39
The slide uses a process flow diagram to illustrate the integration of NVIDIA hardware acceleration into the Databricks stack.present_solution
Open slide detailBeat · Propose Solution
40
The slide illustrates a value-add proposition where NVIDIA software accelerates standard cloud data pipelines.present_solution
Open slide detailBeat · Propose Solution
41
The slide uses a circular inset image showing hardware and software layers, and a process flow diagram for the AI pipeline.present_framework
42
The diagram illustrates a standard two-stage recommendation pipeline (retrieval/candidate generation followed by ranking).present_framework
Open slide detailLoop · Zoom In
43
The slide contrasts legacy CPU-based processing times with GPU-accelerated times while detailing the Merlin software stack.present_solution
Open slide detailLoop · Zoom In
44
analyze_data
Open slide detailLoop · Zoom In
45
The slide illustrates a technical architecture or workflow integration between two NVIDIA products.present_solution
Open slide detailLoop · Zoom In
46
Likely a visual placeholder or demonstration of 3D modeling capabilities.filler
47
The slide shows a bidirectional flow between the human and the avatar, with Jarvis acting as the processing engine (ASR, NLU, TTS, A2F) feeding into Omniverse.present_solution
48
Likely a mascot or branding element used as a transition or filler slide.filler
49
The slide illustrates a technical stack for conversational AI, highlighting the integration of various models (NLU, speech, vision) into a single framework.present_solution
50
The slide illustrates a pipeline: Pre-trained models -> Re-train (NVIDIA AI Toolkit) -> NVIDIA Jarvis (Triton Inference Server) -> Application.present_solution
51
The slide uses a horizontal layout with circular imagery for each industry segment, followed by a logo grid at the bottom.establish_context
52
The slide shows a circular inset of the hardware/software stack on the left and a process flow diagram on the right.present_framework
53
The slide uses visual metaphors of server racks to represent scale and workload diversity.establish_context
54
Product launch slide featuring a high-fidelity render of the hardware and circular icons representing technical specifications.present_solution
55
Features 5 circular icons representing technical capabilities (54B transistors, 3rd Gen Tensor Cores, Sparsity Acceleration, MIG, 3rd Gen NVLink & NVSwitch) andpresent_solution
56
The slide illustrates the bit-width composition of different floating-point formats used in NVIDIA Tensor Cores.present_framework
57
The slide uses a visual metaphor of data cubes to represent precision and range differences.present_solution
58
The slide uses a visual process flow to explain the transition from dense to sparse matrix computation.present_solution
59
The chart shows V100 performance metrics (FP64, FP32, FP16, INT8). The A100 hardware is shown on the right.present_solution
Open slide detailLoop · Before After
60
The chart compares V100 and A100 performance metrics (Peak vs Measured) across FP64, FP32/TF32, FP16, and INT8.quantify_opportunity
Open slide detailLoop · Before After
61
The chart compares V100 and A100 performance metrics, including sparse compute capabilities.quantify_opportunity
Open slide detailLoop · Before After
62
The chart uses a grouped bar structure to compare V100 and A100 performance, highlighting the 'sparse' capability of the A100.analyze_data
Open slide detailLoop · Before After
63
The diagram on the left visually represents the partitioning of a monolithic GPU resource into seven smaller, independent instances.present_solution
Open slide detailBeat · Thesis Headline
64
The slide uses a bar chart to demonstrate performance improvements in AI workloads.analyze_data
Open slide detailBeat · Thesis Headline
65
The slide uses a collage of bird photos as a background to illustrate the audio classification task.analyze_data
Open slide detailBeat · Thesis Headline
66
The slide uses a series of circular icons to represent technical specifications alongside a high-quality product render.front_matter
67
front_matter
68
The slide uses a callout-style diagram to label components of the server hardware.other
69
The slide features a high-quality 3D render of the hardware.present_solution
70
The slide uses a product render to emphasize the hardware architecture.present_solution
71
The slide uses a product render to emphasize the hardware architecture.present_solution
72
The slide uses a logo-grid to establish market credibility and ecosystem support.establish_context
73
analyze_data
74
The slide uses a visual representation of a server rack to anchor the technical specifications and efficiency metrics.quantify_opportunity
75
establish_context
77
The slide uses a high-impact metric (52 Billion Edges / Sec) overlaid on a photo of a data center.illustrate_case
78
The slide highlights a 13x performance improvement and 1/75th cost reduction compared to a baseline (implied).illustrate_case
79
The image shows a large server farm, emphasizing the scale of the infrastructure required for the described data processing.illustrate_case
80
The slide features a high-quality render of a server rack, likely representing the DGX A100 cluster mentioned in the text.illustrate_case
81
The slide uses a high-fidelity 3D render of a data center floor as the background.introduce_nominees
82
The slide uses a high-angle photograph of a data center as a background for the text overlay.quantify_opportunity
83
The slide uses a product render to emphasize the hardware architecture.present_solution
84
The slide uses a horizontal arrangement of icons to represent the breadth of the smart revolution.establish_context
85
The slide uses callouts to label specific components of the hardware shown in the center.present_solution
86
The slide uses a process flow diagram to illustrate how software components (NGC, AI Toolkit) integrate into the EGX stack, culminating in edge hardware.present_solution
87
The slide illustrates a technical stack integration (NVIDIA EGX, Aerial, Metropolis, Deepstream, Jarvis).present_framework
88
The slide illustrates a bidirectional data flow between a physical environment and a digital twin, powered by the NVIDIA EGX hardware stack.present_solution
89
This appears to be a frame from a simulation or promotional video for warehouse automation technology.illustrate_case
90
The slide uses a 'problem-solution' framing by contrasting the extreme complexity of BMW's production (the 'problem') with the specific robotic solutions (the 'illustrate_case
91
The slide uses a hub-and-spoke visual metaphor with the NVIDIA hardware at the center, surrounded by vertical application pillars and categorized partner logos.establish_context
92
The slide uses a horizontal progression to show scalability across different vehicle autonomy levels.present_solution
93
The slide uses a process-flow layout to illustrate the lifecycle of AV development.present_framework
94
This appears to be a high-fidelity 3D render used as a visual transition or background slide.filler
95
The slide uses a logo-grid layout to demonstrate the breadth of the NVIDIA DRIVE partner network.establish_context
96
Visual layout uses a horizontal sequence of product icons and labels.summarize
97
front_matter