A High-Performance Scalable Graphics Architecture

Daniel R. McLachlanDirector, Advanced Graphics Engineering

SGI

Growth in Model Sizes

Worldwide Production of Information

0

20

40

60

80

100

120

140

160

180

200

1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

Exa

byt

es

Source: Gartner

Images courtesy of Parametric Technology Corporation; Photodisc, and Magic Earth, LLC

Problems Are Getting Increasingly Complex Over Time

Bumper Bumper, hood, engine, wheels Entire car

Crash dummyOrgan damage

E-crash dummy

Images courtesy of EAI; SCI Institute, NLM, Theoretical Biophysics Group of

the Beckman Institute at UIUC; Livermore Software Technology Corporation  

The Complexity of the Simple

Images courtesy of Procter & Gamble

Potato Chips

Diapers

2000 2001 2002 2003 2004

Ban

dwid

th S

peci

ficat

ion

(nor

mal

ized

)

Polygons Fill Rate Internal Bus Network I/C

Graphic Cards Are Outpacing PC Architecture and Bandwidth

Performance Gap

Graph based on relative scale.

Perf

orm

ance

2003

Visualization

• Low cost• Fast simple polygons• Single screen image quality

1992

• Extreme resolution• Absolute visual quality• VAN

• Solving complex problems• Dense data sets

Visualization Breaks The Cognitive Barrier For Better DecisionsVisualization Breaks The Cognitive Barrier For Better Decisions

Addressing Real Needs

Images courtesy of Advantage CFD; SCI institute; NLM; Theoretical Biophysics Group of the Beckman Institute at UIUC; Laboratory for Atmospheres, NASA Goddard Space Flight Center; Donghoon Shin, Art Center College of Design, Nvidia Corporation; ATI Technologies, Inc; and Nintendo Co., Ltd.

Clusters Graphics

Cluster Comparison

Pros• Cheap• Industry standard • High display list performance• Good for “embarrassingly parallel”

problems• Can potentially scale to 1000s of

processors

Cons• Cumbersome to program • High administration costs• Few applications for visualization• Difficult to scale for large problems• Difficult to dynamically load balance• Lack of software productivity tools• Often requires data replication• Reliability• Limited to 2GB memory space

Traditional Clusters SGI® NUMAflex™

node+

OS...

Fast NUMAflex™ interconnectGlobal shared memory

node+

OS

...

Commodity interconnectmemmemmemmemmemmem

node+

OS

node+

OS

node+

OS

node+

OS

node+

OS

node+

OS

node+

OS

node+

OS

What is shared memory?• All nodes operate on one large shared memory space, instead of each node having its own

small memory space Shared memory is high-performance• All nodes can access one large memory space efficiently, so complex communication and data

passing between nodes aren’t needed• Big data sets fit entirely in memory; less disk I/O is needed

Shared memory is cost-effective and easy to deploy • It requires less memory per node, because large problems can be solved in big shared memory• Simpler programming means lower tuning and maintenance costs

The Benefits of Shared Memory

1-2 CPUs per node < 64 CPUs per node

How SGI® Onyx® Enables the Role

System at a GlanceScalable Graphics I/O

Scalable Disk I/O Scalable Resolution

Appropriate Delivery

Scalable Rendering

Scalable Data

Compositor

Network

Scalable

Graphics

Scalable

Compute

and

Large

Memory

SGI Onyx

Large

Data

Sets

Scalable Interaction

Moving from a fixed rendering path…

Images courtesy of Pratt and Whitney Canada and Magic Earth, LLC

…to a scalable and programmable rendering path.

Application accelerators

Geometry

Silicon Graphics® Onyx4™ UltimateVision™ Changing the Application Paradigm

ScalingA Shift in Pipe Paradigm

3. Time-based decomposition

Even more powerful in combination

All modes can be used separately orcombined in any number of ways

Even more powerful in combination

All modes can be used separately orcombined in any number of ways

Data courtesy of DaimlerChrysler, Images courtesy of MAK Visible Human public data set

1. Screen-based decomposition

2. Eye-based decomposition

4. Data-based decomposition

Multi-Tier CompositionComposite output of multiple compositors e.g., first layer does 2D composition, second layer does anti-aliasing

Visual ServingComposited output sent to workstations for viewing and/or editing

Compositor Flexibility

Silicon Graphics® Onyx4™ UltimateVision™ System Architecture

2 Graphics Pipes2 Graphics Pipes

CPUCPU

CPUCPU

CPUCPU

CPUCPU

8GB RAM8GB RAM

Memory Controller

Memory Controller

SG

I® N

UM

A s

cala

bili

ty

Standard I/Oor

2 Graphics Pipes

Optional

Silicon Graphics® Onyx4™ UltimateVision™Solving bigger and more complex problems

• World’s most scalable visualization system•Up to 32 GPUs in an SSI architecture

• World-leading computational capability•Up to 64 CPUs per node, scalable to 1024 processors

• Solves system b/w limitations of PCs and clusters•Up to 8 NUMAlink 3 connections to a single shared memory pool

• New-generation programmable graphics architecture•OpenGL Shading Language

Conclusion