NVIDIA QUADRO RTX NVIDIA TURING GPU Turing SM RT Cores Turing SM - - PowerPoint PPT Presentation

▶
nvidia quadro rtx nvidia turing gpu
SMART_READER_LITE
LIVE PREVIEW

NVIDIA QUADRO RTX NVIDIA TURING GPU Turing SM RT Cores Turing SM - - PowerPoint PPT Presentation

Mar 14, 2024 214 likes •513 views

NVIDIA QUADRO RTX NVIDIA TURING GPU Turing SM RT Cores Turing SM RT Cores Up to 10 Giga Rays/sec Up to 16 TFLOPS + 16 TIPS Ray Triangle Intersection Concurrent FP & INT Execution BVH Traversal Unified L1 Cache Variable Rate Shading

slide-1
SLIDE 1

NVIDIA QUADRO RTX

slide-2
SLIDE 2

2

Display Display

Native HDR 8K DisplayPort Virtual Link

NVLink NVLink

Up to 100 GB/sec GPU-GPU Memory Access

Video Video

HEVC 8K Real Time Encode 25% Improved Bitrate

Memory Memory

6MB L2 Cache Up to 384-bit GDDR 6 @ 14Gbps Up to 672 GB/sec

Turing SM Turing SM

Up to 16 TFLOPS + 16 TIPS Concurrent FP & INT Execution Unified L1 Cache Variable Rate Shading

RT Cores RT Cores

Up to 10 Giga Rays/sec Ray Triangle Intersection BVH Traversal

Tensor Cores Tensor Cores

Up to 130 TFLOPS FP16 Up to 260 TOPS INT8 Up to 500 TOPS INT4

NVIDIA TURING GPU

slide-3
SLIDE 3

3

TURING FOR PROFESSIONAL WORKFLOWS

RT CORES

Brings real-time ray tracing to professional graphics workflows

TENSOR CORES

Enables AI-augmented tools and applications

ADVANCED SHADERS

Powers next-generation of graphics, VR, and GPU compute workflows

slide-4
SLIDE 4

4

WHAT IS RAY TRACING? WHAT IS RAY TRACING?

  • Models the behavior of

light in the scene

  • Produces accurate model
  • f the real world –

photorealistic images

  • Computationally intensive
slide-5
SLIDE 5

5

TURING RT CORES TURING RT CORES

Hardware Accelerated Tracing of Rays Through the Scene Key Benefits:

  • Real-time ray tracing in the

application viewport allows for instantaneous feedback and review iteration

  • Accelerated offline rendering lets

you create photorealistic images faster

  • Make better decisions, faster,

more iterations without impacting schedules

slide-6
SLIDE 6

6

TURING TENSOR CORES TURING TENSOR CORES

Next Generation of Hardware Accelerated Deep Learning Key Benefits:

  • Turing Tensor Cores deliver fast

inferencing performance and support additional precision modes, which boosts inferencing workload performance

  • Bring new techniques like Deep

learning Super Sampling (DLSS) to your workstation via hardware- accelerated deep learning enabled tools and applications

slide-7
SLIDE 7

7

TURING ADVANCED SHADERS TURING ADVANCED SHADERS

Advanced Graphics Technology Key Benefits:

  • Create more objects per scene with

more flexible control over the level of detail

  • Finer control over shading allows for

more dynamic geometry manipulation, letting developers deploy new,

  • ptimized algorithms
  • Enhancements to single-pass stereo

provide greater flexibility and support for new generation of HMDs

Shading Re-Use Foveated Shading Mesh Shading to render thousands

  • f objects in real time

200º FOV HMD using MVR

slide-8
SLIDE 8

8

TURING VR TURING VR

VISUAL QUALITY ULTRA WIDE FIELD OF VIEW ACOUSTIC SIMULATION EASY SETUP

Turing GPU features enhance VR Key Benefits:

  • Optimize resolutions with variable

rate shading and foveated rendering

  • Multi-view rendering provides a

wider field of view and support for next-gen HMDs & displays

  • RT Cores enable accurate acoustic

simulations to deliver more realistic virtual environments

  • Easier set up with VirtualLink™

single cable connection

slide-9
SLIDE 9

9

QUADRO RTX VIRTUALLINK™ *

VirtualLink

*In preparation for the emerging VirtualLink standard, Turing GPUs have implemented hardware support according to the “VirtualLink Advance Overview”. To learn more about VirtualLink, please see http://www.virtuallink.org.

VirtualLink USB-C Port

VirtualLink is an industry standard Alternate Mode of USB Type-C™ designed to deliver the power, display, and data required to power VR headsets through a single USB Type-C connector.

  • 4 lanes HBR3 DisplayPort
  • USB 3.1 Gen2 SuperSpeed
  • 27 W power
  • Industry consortium includes: NVIDIA, VALVE,

Oculus, AMD, Microsoft

  • virtuallink.org for more details
slide-10
SLIDE 10

10

QUADRO RTX NVLINK QUADRO RTX NVLINK

M E M O R Y PERFORMANCE

2x RTX 5000 32 GB 2x RTX 6000 48 GB 2x RTX 8000 96 GB

High-speed GPU interconnect Key Benefits:

  • Scaled memory and performance

lets you split workloads efficiently across two GPUs, sharing up to 96 GB of memory capacity

  • Increased bandwidth enables new,

advanced SLI display topologies that were previously impossible with PCIe-based solutions

*application support for NVLink required

slide-11
SLIDE 11

11

QUADRO NVLINK

Quadro Family NVLink Bridges

Quadro GPU NVLink Bridge Slot Configuration Bandwidth Bridges Required Quadro RTX 8000 RTX 6000 Quadro RTX NVLink HB 2- Slot 2-Slot Up to 100 GB/s 1 Quadro RTX NVLink HB 3- Slot 3-Slot Quadro RTX 5000 Quadro RTX NVLink 2-Slot 2-Slot Up to 50 GB/s 1 Quadro RTX NVLink 3-Slot 3-Slot Quadro GV100 NVLink GV100 2-Slot Up to 200 GB/s 2 Quadro GP100 NVLink GP100 2-Slot Up to 160 GB/s 2

Bridges are product specific, not cross-compatible RTX 6000 RTX 5000 GV100 GP100

*not final product images

Quadro RTX boards only require 1 NVLink bridge GV100/GP100 boards require 2 NVLink bridges

RTX 8000

slide-12
SLIDE 12

12

QUADRO RTX FOR AI

Pro Applications Inferencing Aggregation Inferencing At-The-Edge Quadro RTX ideal for AI augmented professional applications and professional AI inferencing deployments

slide-13
SLIDE 13

13

QUADRO RTX FOR AI - NGX

The NVIDIA NGX SDK makes it easy for developers to integrate AI features into their applications with pre- trained neural networks. NGX provide AI-augmented features for video and image processing including:

  • AI InPainting

Allows the removal of existing content from images and replaces it with realistic computer-generated alternatives.

  • AI Up-Res

Increases the resolution of an image or video by 2x, 4x or 8x using AI to create new pixels by interpreting the image & intelligently placing data in the new image.

  • DLSS: (Deep Learning Super Sample)

Removes jagged lines to smooth images, producing a higher quality image faster than by using other techniques.

  • AI Slow-Motion

Inserts interpolated frames into a video stream to provide smooth, slow-motion video Details on the NGX SDK: developer.nvidia.com/rtx/ngx

AI InPainting AI Up-Res DLSS AI Slow-Motion

NGX AI-based features

slide-14
SLIDE 14

14

QUADRO RTX VALUE FOR INDUSTRIES

RENDERING

Content Creation Product Design Building Design

AI

Up Res Generative Design Generative Design

VR

Design Review Content Creation Design Review

MEDIA & ENTERTAINMENT MANUFACTURING AEC

Real-time rendering speeds up the creative workflow AI-augmented tools accelerate the creative process VR powers design reviews, compelling content creation and entertainment experiences

slide-15
SLIDE 15

QUADRO RTX

RTX 6000 RTX 5000 RTX 4000

slide-16
SLIDE 16

16

GPU Architecture Turing CUDA Cores 4608 RT Cores 72 Tensor Cores 576 Memory Size 24 GB GDDR6 Memory BW Up to 672 GB/s NVLink 2-way (2 & 3slot) 100 GB/s bidirectional Display Support 4x DP + 1x VirtualLink VR Ready Yes VirtualLink™ Yes Advanced Display SYNC 2 Board Power Total Board Power: 295W Total Graphics Power: 260W Power Connectors 1x 8-pin, 1x 6-pin PCIe

QUADRO RTX 6000 KEY SPECIFICATIONS

slide-17
SLIDE 17

17

UPGRADING TO RTX 6000

RTX 6000

P6000 M6000 24GB Benefit

Architecture Turing Pascal Maxwell Latest generation NVIDIA GPU technology CUDA Cores 4608 3840 3072 Fast graphics and compute performance RT Cores 72

  • GPU accelerated ray tracing for interactive and

batch rendering Tensor Cores 576

  • GPU accelerated Deep Learning for AI-augmented

applications Memory 24 GB GDDR6 Up to 672 GB/s 24 GB GDDR5X

Up to 432 GB/s

24 GB GDDR5

Up to 317 GB/s

Smooth interaction with complex models, faster render & compute performance NVLink 2-way

  • Scales memory & compute up to 48 GB for largest

renders, models and datasets VR Ready Multi-View Rendering Single pass stereo Yes Latest generation of GPU accelerated immersive VR technology VirtualLink Yes

  • Simplified single cable VR HMD connectivity
slide-18
SLIDE 18

18

RTX 6000 UP TO 2X FASTER THAN PREVIOUS GENERATION*

*based on M6000 SPECviewperf 13 performance Test run on a workstation with Xeon Gold 6154 3GHz (3.7 GHz turbo). 64GB RAM, Windows 10 64-bit, NVIDIA driver version 341.49 & 411.61. Performance testing completed with publicly available SPECviewperf 13 benchmark information.

1.69 1.42 1.34 1.65 1.82 1.71 2.05 2.12 2.00 1.34 Geomean 3dsmax-06 catia-05 creo-02 energy-02 maya-05 medical-02 showcase-02 snx-03 sw-04 0.00 0.50 1.00 1.50 2.00 2.50

SPECviewperf 13

Relative Performance

M6000 P6000 RTX 6000

slide-19
SLIDE 19

19

RTX 6000 MORE THAN 3X FASTER THAN COMPETITION*

*based on Radeon Pro WX9100 SPECviewperf 13 performance Test run on a workstation with Xeon Gold 6154 3GHz (3.7 GHz turbo). 64GB RAM, Windows 10 64-bit, NVIDIA driver version 411.61, AMD driver version 18.Q4. Performance testing completed with publicly available SPECviewperf 13 benchmark information.

1.67 1.28 1.38 1.78 3.39 1.51 1.34 1.59 2.24 1.30 geomean 3dsmax-06 catia-05 creo-02 energy-02 maya-05 medical-02 showcase-02 snx-03 sw-04 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00

SPECviewperf 13

Relative Performance

WX9100 RTX6000

slide-20
SLIDE 20

20

GPU Architecture Turing CUDA Cores 3072 RT Cores 48 Tensor Cores 384 Memory Size 16 GB GDDR6 Memory BW Up to 448 GB/s NVLink 2-way (2 & 3slot) 50 GB/s bidirectional Display Support 4x DP + 1x VirtualLink VR Ready Yes VirtualLink™ Yes Advanced Display SYNC 2 Board Power Total Board Power: 265W Total Graphics Power: 230W Power Connectors 1x 8-pin, 1x 6-pin PCIe

QUADRO RTX 5000 KEY SPECIFICATIONS

slide-21
SLIDE 21

21

UPGRADING TO RTX 5000

RTX 5000

P5000 M5000 Benefit

Architecture Turing Pascal Maxwell Latest generation NVIDIA GPU technology CUDA Cores 3072 2560 2048 Fast graphics and compute performance RT Cores 48

  • GPU accelerated ray tracing for interactive and

batch rendering Tensor Cores 384

  • GPU accelerated Deep Learning for AI-augmented

applications Memory 16GB GDDR6 Up to 448 GB/s 16 GB GDDR5X

Up to 288 GB/s

8 GB GDDR5

Up to 211 GB/s

Smooth interaction with complex models, faster render & compute performance NVLink 2-way

  • Scales memory & compute up to 48 GB for largest

renders, models and datasets VR Ready Multi-View Rendering Single pass stereo Yes Latest generation of GPU accelerated immersive VR technology VirtualLink Yes

  • Simplified single cable VR HMD connectivity
slide-22
SLIDE 22

22

RTX 5000 MORE THAN 2X FASTER THAN PREVIOUS GENERATION*

*based on M5000 SPECviewperf 13 performance Test run on a workstation with Xeon Gold 6154 3GHz (3.7 GHz turbo). 64GB RAM, Windows 10 64-bit, NVIDIA driver version 341.49 & 411.61. Performance testing completed with publicly available SPECviewperf 13 benchmark information.

Relative Performance

1.84 1.45 1.50 1.84 2.02 1.92 2.22 2.19 2.39 1.31 Geomean 3dsmax-06 catia-05 creo-02 energy-02 maya-05 medical-02 showcase-02 snx-03 sw-04 2.50 2.00 1.50 1.00 0.50 0.00

SPECviewperf 13

Relative Performance

M5000 P5000 RTX 5000

slide-23
SLIDE 23

23

RTX 5000 UP TO 2X FASTER THAN COMPETITION*

*based on Radeon Pro WX8200 SPECviewperf 13 performance Test run on a workstation with Xeon Gold 6154 3GHz (3.7 GHz turbo). 64GB RAM, Windows 10 64-bit, NVIDIA driver version 411.61, AMD driver version 18.Q4. Performance testing completed with publicly available SPECviewperf 13 benchmark information.

1.38 1.08 1.26 1.60 2.23 1.31 1.12 1.12 1.88 1.23 geomean 3dsmax-06 catia-05 creo-02 energy-02 maya-05 medical-02 showcase-02 snx-03 sw-04 0.00 0.50 1.00 1.50 2.00 2.50

SPECviewperf 13

Relative Score

WX8200 RTX5000

slide-24
SLIDE 24

24

GPU Architecture Turing CUDA Cores 2304 RT Cores 36 Tensor Cores 288 Memory Size 8 GB GDDR6 Memory BW Up to 416 GB/s NVLink N/A Display Support* 3x DP + 1x VirtualLink VR Ready Yes VirtualLink™ Yes Advanced Display SYNC 2 Board Power Total Board Power: 160W Total Graphics Power: 125W Power Connectors 1x 8-pin

QUADRO RTX 4000 KEY SPECIFICATIONS

*RTX 4000 can support 4x DP 1.4 using the included USB-C to DisplayPort adapter

slide-25
SLIDE 25

25

UPGRADING TO RTX 4000

RTX 4000

P4000 M4000 Benefit

Architecture Turing Pascal Maxwell Latest generation NVIDIA GPU technology CUDA Cores 2304 1792 1664 Fast graphics & compute performance RT Cores 36

  • GPU accelerated ray tracing for interactive and

batch rendering Tensor Cores 288

  • GPU accelerated Deep Learning for AI-augmented

applications Memory 8GB GDDR6 Up to 416 GB/s 8 GB GDDR5

Up to 288 GB/s

8 GB GDDR5

Up to 192 GB/s

Smooth interaction with complex models, faster render & compute performance VR Ready Multi-View Rendering Single pass stereo No Latest generation of GPU accelerated immersive VR technology VirtualLink Yes

  • Simplified single cable VR HMD connectivity
slide-26
SLIDE 26

26

RTX 4000 UP TO 3X FASTER THAN PREVIOUS GENERATION*

*based on M4000 SPECviewperf 13 performance Test run on a workstation with Xeon Gold 6154 3GHz (3.7 GHz turbo). 64GB RAM, Windows 10 64-bit, NVIDIA driver version 341.49 & 411.61. Performance testing completed with publicly available SPECviewperf 13 benchmark information.

2.23 1.89 1.95 2.44 2.31 2.34 3.02 2.47 2.77 1.38 Geomean 3dsmax-06 catia-05 creo-02 energy-02 maya-05 medical-02 showcase-02 snx-03 sw-04 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50

SPECviewperf 13

Relative Performance

M4000 P4000 RTX4000

slide-27
SLIDE 27

27

RTX 4000 1.8X FASTER THAN COMPETITION*

*based on AMD WX7100 SPECviewperf 13 geomean score Test run on a workstation with Xeon Gold 6154 3GHz (3.7 GHz turbo). 64GB RAM, Windows 10 64-bit, NVIDIA driver version 411.61, AMD driver version 18.Q4. Performance testing completed with publicly available SPECviewperf 13 benchmark information.

1.84 1.43 1.60 1.92 6.07 1.72 1.42 1.55 1.86 1.31 geomean 3dsmax-06 catia-05 creo-02 energy-02 maya-05 medical-02 showcase-02 snx-03 sw-04 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00

SPECviewperf 13

Relative Performance

WX7100 RTX4000

slide-28
SLIDE 28

28

RTX 4000 THE BEST PROFESSIONAL GRAPHICS CARD UNDER $1000

*based on AMD WX8200 SPECviewperf 13 geomean score Test run on a workstation with Xeon Gold 6154 3GHz (3.7 GHz turbo). 64GB RAM, Windows 10 64-bit, NVIDIA driver version 411.61, AMD driver version 18.Q4. Performance testing completed with publicly available SPECviewperf 13 benchmark information.

1.84 1.43 1.60 1.92 6.07 1.72 1.42 1.55 1.86 1.31 geomean 3dsmax-06 catia-05 creo-02 energy-02 maya-05 medical-02 showcase-02 snx-03 sw-04 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00

SPECviewperf 13

Relative Performance

WX7100 WX8200 RTX4000

slide-29
SLIDE 29