NVIDIA

1
NVIDIA MAXIMUS FOR MANUFACTURING

• CREATE WITHOUT THE WAIT

2
The Vision Reality-Based Design
Real-World Materialsduring Interactive Design
Real-World Physics during Interactive Design

• Global Lighting
• Physical Properties
• Photorealism Light Simulation

• Mechanical Dynamics
• Fluid Dynamics
• Structural Dynamics

3
Reality-Based Design Today

• REAL-WORLD PHYSICS

REAL-WORLD MATERIALS
REAL-TIME INTERACTIVITY
4
Traditional Product Development Steps
Compute-Intensive Applications Migrated from
Workstations to CPU Clusters Design Iterations
with Separate Solutions and Groups are Time and
Cost Adders
5
Traditional Workstation
Design or Simulate. Design or Render. Never Both.
OpenGL / DX GPU
SystemCPU

6
The New Workstationwith Maximus Technology
Design GPU
SystemCPU
Simulate GPU
7
Reinventing Workflow
Design
Simulate (CPU)
Simulation Time y
Design Time x
Traditional Workstation
Design Time x
Design

Simulate
NVIDIA MAXIMUS Workstation
Simulation Time 1/10 y
8
Reinventing Workflow
Design
Simulate (CPU)
Simulation Time y
Design Time x
Traditional Workstation
Design Time x
Design
Design2
Design3
Design4




Simulate
Simulate2
Simulate3
Simulate4
NVIDIA MAXIMUS Workstation
Simulation Time 1/10 y
9

• Intelligent GPU job allocation
• A Single Unified Driver
• Full ISV Application Certifications
• A range of OEM Workstations

NOVEMBER 14, 2011
10
NVIDIA MAXIMUS Powered Rendering
Downstream UseImages Movies
Styling / Industrial Design
11
Typical Styling/Design Process
Styling Development Steps with Offline
Photorealistic Rendering (Slow)
Design 1c
Design 1a
Design 1b
Render
Render
Render
Review 1a
Review 1c
Review 1b
Surface

• Design reviews wait for image and animation
  rendering
• Short-notice executive reviews are a frequent
  challenge
• Design-revision loops increase time and cost

12
NVIDIA MAXIMUS Powered Design/Styling
Styling Development Steps with Real-time
Photorealistic Rendering
Design 1c
Design 1a
Design 1b
Review 1a
Review 1c
Review 1b
Class-A Surface Creation
No separate rendering steps!
13
Customer Success Story Mercedes Benz Concept Car
Design

• Challenge
• Render hundreds of frames per animated sequence
• Traditionally took 10 min/frame, people left
  idle, rendering had to be done at night
• Compromises made as not all designs could be
  finished within a week
• Solution
• NVIDIA Maximus with Quadro 6000 and Tesla C2075
  running BunkSpeed, Maya, Alias, 3DSMax
• Impact
• Each frame now rendered in less than 1 min
• While rendering, work on next sequence continues,
  significantly increasing productivity
• No more compromising

14
Customer Success Story CLEAT, Inc.

• Challenge
• Rendering of 360 degree model movies took
  overnight to complete
• Took 10 min to change wheel color
• Separate systems used for viewing data
  rendering (long wait times for data downloads and
  uploads)
• Solution
• NVIDIA Maximus workstation with Tesla C2075
  
• Quadro 5000 running Autodesk 3ds Max with
  iray
• NVIDIA Maximus Impact
• 360 degree rotating movies now produced in 30 min
  ready for customer viewing
• Rendering time decreased from 10 min to 30
  seconds, improving workflow design review
  process
• Quicker time to higher quality results mean
  customers place more orders

15
Customer Success Story Fisker

• Challenge
• Still frame rendering took 1-20 min
• Animations took several hours, often overnight
• Rendering monopolized system resources
• Ideas were lost as designers could not act on
  them
• Solution
• NVIDIA Maximus with Quadro 6000 and Tesla C2075
  running Bunkspeed, CATIA, Alias
• Impact
• Rendering time reduced by 5X
• Design work continues simultaneously with
  rendering and more designs are acted upon

16
NVIDIA MAXIMUS Powered Simulation
Core Engineering
17
Double the Local Simulation Performance
CPU vs CPU GPU More than 2x the Simulation
Perf. 2x the Design Candidates. Faster
Interactive Design Perf.
CPU GPU
Design 2a
Design 1c
Design 1a
Design 2b
Design 1b
Simulate 1a
Simulate 2a
Simulate 1b
Simulate 2b
12 Cores
6 Cores Quadro GPU
Design 4b
Design 3b
Design 4a
Design 3a
Design 2c
Design 1c
Design 2b
Design 1b
Design 2a
Design 1a
Simulate 1a
Simulate 2a
Simulate 1b
Simulate 2b
Simulate 1c
Simulate 2c
Simulate 3a
Simulate 4a
Simulate 3b
6 Cores Tesla GPU
18
Customer Success Story Astrobotic

• Challenge
• Computationally intensive engineering left
  workstations as dead zones engineers idle
• Analysis for each what if scenario took 30 min
• File sizes had to be restricted
• Solution
• NVIDIA Maximus with Quadro 6000 and Tesla C2075
  SolidWorks, ANSYS, Matlab
• Impact
• What if scenarios reduced to 15 min
• Test models more thoroughly due to increased file
  size capability
• Additional projects possibilities can now be
  explored while computations are running

19
Typical Small Medium Businesses
In-house Design and Engineering. Outsourced
Simulation.
Review 1a
Design 1c
Design 1a
Review 1c
Design 1b
Review 1b
Render
Render
Render
Delay
Simulate 1a
Delay
Simulate 1b
Delay
Simulate 1c
Outsourced

• Design reviews wait for image and animation
  rendering
• Simulation has to be outsourced to keep
  workstations available

20
Typical Small Medium Businesses
Local Simulation Removes Delay. Adds Confidence
for No Prototype!
Review 1a
Design 1c
Design 1a
Review 1c
Design 1b
Review 1b
Simulate 1a
Simulate 1b
Simulate 1c
Straight to Production
Everything done in-house without losing
productivity. Often no prototypes required!
21
Customer Success Story Liquid Robotics

• Challenge
• Computational simulation needed to increase
  number of design options
• Reduce number of physical prototypes
• Simulations had to be run at night or outsourced
• Solution
• NVIDIA Maximus with Quadro 6000 and Tesla C2075
  running SolidWorks, ANSYS, Matlab
• Impact
• Work continues on same systems while simulations
  are running
• 4x increase in mechanical design options
  available for review
• Higher quality final design reached before
  physical prototype

22
Entry to CAE on GPU
Price-Performance
Max Performance

• Single CPU
• 32 GB RAM
• One NVIDIA Quadro 2000 GPU
• One NVIDIA Tesla C2075 GPU

• Dual CPUs
• 32 48 GB RAM
• One NVIDIA Quadro 4000 GPU
• One NVIDIA Tesla C2075 GPU

• Dual CPUs
• 48 GB RAM (or more)
• One NVIDIA Quadro 6000 GPU
• One NVIDIA Tesla C2075 GPU

Memory requirements dictated by problem size to
minimize disk I/O.
23
Benefits Summary
GPU Horsepower
10X
CPU Horsepower
Higher Performance
Efficient Multi-Tasking
ISV Certified

• Tesla Quadro board
• certifications
• Global OEM system
• certifications
• All leading CAD CAE ISVs

• Up to 4x faster simulation
• Up to 9x faster rendering
• 100 interactive designperformance while
  renderingor simulating

• Design AND Simulate
• Design AND Render
• Seamless Experience

24
How to Get Started
25
Backup Benchmark Slides
26
Rendering Performance NVIDIA iRayPowered by
NVIDIA Maximus Technology

• CATIA V6 LiveRendering
• Autodesk 3ds Max ActiveShade
• Bunkspeed Shot Move
• Enabled by NVIDIA iRay
• Accelerated by Maximus
• Physically Correct Rendering
• Easy to use!

27
Rendering Performance NVIDIA iRayPowered by
NVIDIA Maximus Technology

• CATIA V6 LiveRendering
• Autodesk 3ds Max ActiveShade
• Bunkspeed Shot Move
• Enabled by NVIDIA iRay
• Accelerated by Maximus
• Physically Correct Rendering
• Easy to use!

28
Superior Performance
NVIDIA Quadro 5000 GPU
Quadro FX 4800
Quadro FX 4600
3.5x
Quadro 4000
Quadro FX 3800
Quadro FX 3700
3.3x
Quadro 2000
Quadro FX 1800
Quadro FX 1700
3.2x
Results obtained by running industry accepted
standard graphics benchmark tests (SPECViewperf
11) for SolidWorks, comparing Quadro 4000 or 2000
to Quadro FX 1700.  The comparison is based on
standard industry workstations representative of
when each particular graphics product was
launched. n Resolution of 1900 x 1200, Windows
Win 7, 64 Bit, System 1P XEON 3.3GHz, 6GB Memory.
n-2 Resolution of 1900 x 1200, WinXP x64, 1x
3GHz Core2 Quad (QX9650) 4GB.
29
ANSYS and SIMULIA/AbaqusMore than Twice the
Design Candidates. Up to 4.2x Faster.
Up to 3.5x Faster with NVIDIA Tesla GPU
Benchmark V13sp-5 Model- Turbine Geometry2.1M
DOF, Static Nonlinear, Direct Sparse CPU 2 x
Westmere Xeon 5670 at 2.93 GHz
Benchmark s4b Model- Engine Geometry, 5M
DOFStatic nonlinear, Direct Sparse CPU Westmere
Xeon 5680 at 3.33 GHz
30
NVIDIA MAXIMUS No Loss in SolidWorks
Performance
0 Loss
25 - 53 Loss
31
AND With Faster ANSYS Performance
1.5x Faster
1.9x Faster
24 Loss
Benchmark V13sp5 Model- Turbine Geometry,2.1M
DOF, Static nonlinear, Direct Sparse
32
AND For Less Money!