GraphStream Scalable Computing Platforms
Integrated scalable computer systems, optimized to support the most
challenging technical and visual computing applications in energy,
finance, manufacturing, media, medicine, national security, and
science
Based exclusively on mainstream
open-standard architecture and leadership-class Commercial
Off-The-Shelf (COTS) hardware and software components for processing
(CPU, GPU), storage, networking, power, and cooling
Efficiency superior to currently available alternatives across a wide
range of scales, from a single workstation/server up to thousands of
cluster nodes
Straightforward building-block system configuration using standard
rackmount modules
Superior build quality and attention to detail
World-class systems engineering and support
Strong track record plus non-proprietary pure-COTS architecture equals
minimum supplier risk
Highly competitive pricing
Key application areas
Scalable GPU-accelerated general-purpose computation (GPGPU)
Scalable GPU-accelerated interactive visualization of large data sets
Scalable remote visual application serving
Interactive environments with high-resolution displays
Example larger GraphStream system installations
Argonne National Laboratory "Eureka" system [Details] [Update]
Primary visualization resource
for IBM BlueGene/P leadership-class HPC system, 6/2008 Top500 #4
System
Configuration
214 Intel Xeon quad-core CPUs
212 NVIDIA Quadro GPUs
Myri-10G and Ethernet interconnects
King Abdullah University of Science and Technology [Details]
Primary visualization resource
for IBM BlueGene/P leadership-class HPC system, 6/2009 Top500 #14
System
Configuration
86 Intel Xeon quad-core CPUs
147 NVIDIA Quadro GPUs
96TB InfiniBand-attached storage
InfiniBand-QDR and Ethernet interconnects
Lawrence
Livermore National Laboratory "Gauss" system [Details]
Primary visualization resource
for IBM BlueGene/L leadership-class HPC system, 11/2007 Top500 #1
System
Configuration
512 AMD Opteron CPUs
3 terabytes main memory
256 NVIDIA Quadro GPUs
InfiniBand and Ethernet interconnects
18 47RU rackmount
enclosures
225,000 watt
three-phase power feed
256 separate Gigabit
Ethernet links to external 900-terabyte parallel
clustered storage system
Selected processing/storage modules
P1C6A-TR4B Workstation/Server
Convertible tower / 4RU-rackmount chassis
Optimized for use in quieter environments; 27 dB core-platform
noise level
Up to 8way-SMP via Intel Core i7 quad-core CPU supporting two threads
per core, up to 3.33GHz
Up to 38.4 GByte/sec peak aggregate main-memory bandwidth using
DDR3-1600 modules
Up to 24GB memory (4GB * 6)
Up to 2 internal GPU cards or 4 external GPU cards, 8 dual-link DVI outputs
Up to 26TB internal storage
865W power supply
P2X12B-R1B GPU Computing Server
1RU-rackmount chassis
Up to 16way-SMP via 2pc quad-core Intel Xeon 5500 CPUs supporting two
threads per core, up to 2.93GHz
Up to 64 GByte/sec peak aggregate main-memory bandwidth using
DDR3-1333 modules
Up to 96GB memory (8GB * 12)
Up to 2 internal GPU cards or 4 external GPU cards, 8 dual-link DVI outputs
Up to 6TB internal storage
1400W easy-swap power supply
P2X18A-R2A Graphics Server
2RU-rackmount chassis
Up to 16way-SMP via 2pc quad-core Intel Xeon 5500 series CPU
supporting two threads per core, up to 3.2GHz
Up to 64 GByte/sec peak aggregate main-memory bandwidth using
DDR3-1333 modules
Up to 192GB memory (16GB * 12)
Up to 8 external GPU cards, 16 dual-link DVI outputs
Up to 18TB internal storage
720W 1+1 modular redundant hot-swap power supply
P4O32A-R2C Graphics Server
2RU-rackmount chassis
24way-SMP via 4pc 6-core AMD Opteron CPU
Up to 256GB memory (8GB * 32)
Up to 6 external GPU cards, 12 dual-link DVI outputs
Up to 13TB internal storage
1200W 1+1 modular redundant hot-swap power supply
P4O32A-TR4D Workstation/Server
Convertible tower / 4RU-rackmount chassis
24way-SMP via 4pc 6-core AMD Opteron CPU
Up to 256GB memory (8GB * 32)
Up to 2 internal GPU cards or 6 external GPU cards, 12 dual-link DVI outputs
Up to 20TB internal storage
1200W 1+1 modular redundant hot-swap power supply
Each of the above platforms also supports:
10Gbps Ethernet and 32Gbps InfiniBand networking
Advanced standards-based Lights-Out Management via service processor
supporting IPMIv2.0, embedded Web server, dedicated management
Ethernet port
Cost-optimized entry-level configurations
For more information or to request a quote, please call 650-595-5324
or send email to
info@graphstream.com .
Copyright 2009 GraphStream Incorporated. All Rights Reserved.
NVIDIA Tesla GPU Computing Solutions