Tel: +86-755-26639997    Fax: +86-755-26631695
 
 
Skip Navigation Links

SGI® ICE™ X

Petascale Performance Leadership

SGI ICE X is the next generation of the world's fastest distributed memory supercomputer for over four years running. This performance leadership is proven time and again, not just in the lab, but at customer sites including the largest and fastest pure compute InfiniBand cluster in the world.

The system can be configured with compute nodes comprising Intel® Xeon® processor E5-2600 v2 series exclusively or with compute nodes comprising both Intel® Xeon® processor E5-2600 series and Intel® Xeon Phi™ Coprocessors. Running on standard Linux, SGI ICE X can deliver over 172 teraflops per rack and scale from 36 to tens of thousands of nodes to solve the world's most challenging compute problems.

SGI ICE X is designed to minimize system overhead and communication bottlenecks that can rob efficiency and scalability. The system, for example, offers the highest performance and scalability for computer fluid dynamics (CFD), demonstrating up to 3,072 cores processing in parallel.

Production-Ready in Hours or Days, Not Weeks or Months

Power up and go, at scale, in less time and with less effort.

Built entirely on industry-standard hardware and software components, SGI ICE X enables access to the full spectrum of the Linux ecosystem. SGI Performance Suite optimizes the performance of Linux applications. SGI Foundation Suite optimizes drivers and system monitoring. And SGI Management Suite helps ease administration, including all-important power management, and maintain high reliability, availability and serviceability.

Ultimate Flexibility and Seamless Scalability

SGI ICE X is the only system in its class offering expandability within and across technology generations while maintaining uninterrupted production workflow, and is the only platform capable of seamless scalability from tens of teraflops to tens of petaflops. Clusters can be architected in a variety of topologies with choice of switch and single or dual plane Infiniband blade connection. The integrated bladed design offers rack-level redundant power and cooling via air, warm water or cold water, for enhanced reliability and availability. The result is a system with unmatched efficiency, performance and overall value.

Also available with storage and visualization options, SGI ICE X sets a new standard for flexibility, simplicity and ease-of-use in scale-out computing. Along with industry-leading professional services and support, SGI enables customers to efficiently overcome the challenges of complex data intensive workflows and accelerate results.

Compute Blades

IP-113

IP-115

IP-119

Processors

• Intel® Xeon® Processor E5-2600 v2 series

• Intel® Xeon® Processor E5-2600 v2 series

• Intel® Xeon® Processor E5-2600 v2 series
• Intel® Xeon Phi™ Coprocessor

Memory

• 16 DDR3 DIMM slots (8 per CPU socket)
• 4, 8, 16 and 32GB 1866 MT/s ECC Registered DIMMs

• 16 DDR3 DIMM slots per twin blade (4 per CPU socket)
• 4, 8, 16 and 32GB 1866 MT/s ECC Registered DIMMs

• 8 DDR3 DIMM slots (4 per CPU socket)
• 4, 8, 16 and 32GB 1866 MT/s ECC Registered DIMMs

Storage

• Two 2.5" SATA drives (HDD or SSD) per blade

• Two 2.5" SATA drives (HDD or SSD) per twin blade; one per logical node

• Two 2.5" SATA drives (HDD or SSD) per blade

FDR IB Mezzanine Card

• Single Port, Dual Port or Dual Single Port

• Dual Single Port

• Single Port, Dual Port or Dual Single Port

Cooling

• Traditional Heat Sinks

• SGI Cold Sinks

• SGI Cold Sinks

Topology Options

• Single or dual plane all-to-all, fat tree, hypercube and enhanced hypercube

• Single plane all-to-all, fat tree, hypercube and enhanced hypercube

• Single or dual plane all-to-all, fat tree, hypercube and enhanced hypercube

Blade Enclosures

Standard

Premium

Integrated Switch

• Single 36 port FDR IB ASIC with 18 ports external

• Dual 36 port FDR IB ASIC with 48 ports external

Administrative Network

• Dedicated GigE network (redundancy optional), chassis management controller

• Dedicated GigE network (redundancy optional), chassis management controller

Racks

D-Rack
(For IP-113 Compute Blade)

M-Rack
(For IP-115 Compute Blade)

Dimensions

• 24"W x 49.5"D, air cooled
• 24"W x 54.75"D, water cooled (42U Standard)
• Can optionally be extended to 48U

• 32"W x 40"D (42U standard)

Blade Enclosure Support

• Up to two blade enclosure pairs (72 total blade slots)

• Up to two blade enclosure pairs (72 total blade slots)

Power

• Up to 5+1 redundant 3000W power supplies per blade enclosure pair

• Up to 17+1 redundant 3000W power supplies per blade enclosure pair

Cooling

• Open-looped airflow or water (optional)

• Closed loop airflow with warm water support capability

Storage

InfiniteStorage InfiniBand Solutions

• High performance shared file systems
• IP over InfiniBand

• Native InfiniBand block level access
• Native InfiniBand SAN supported with CXFS

System Management

Hierarchical System Management

Tier 1: System Administration Controller

• One per SGI ICE system
• Provisions out software to Rack Leader Controller (RLC)
• Pulls aggregated cluster management data from RLC
• Utilizes C1104-RP7 "6017" system

Tier 2: Rack Leader Controller

• One per two-blade enclosure pair
• Holds blade boot images
• Runs fabric management software
• Aggregates cluster management data for rack
• Utilizes C1104-RP7 "6017" system

Tier 3: Chassis Management Controller

• Two or four per blade enclosure pair
• Controls master power to all compute nodes
• Monitors power and blade enclosure environment



Tier 4: Baseboard Management Controller

• One per compute node
• Controls board-level hardware
• Monitors compute node environment

Service Node Options

• Login Node
• Gateway Node
• Batch Node
• Storage Node
• OSS Node
• MDS Node

Service nodes can be optionally configured with:
• GPUs such as NVIDIA® Quadro® FX, NVIDIA® Quadro® and NVIDIA® Tesla™
• Hard Disk Drives (SAS and/or SATA)
• I/O cards (various)

System Software

Operating Systems

• SUSE® Linux® Enterprise Server 11

• Red Hat® Enterprise Linux 6

Cluster Solution Stack

• SGI Foundation Software 2: Optimized drivers and system monitoring
• SGI Performance Suite optimized application performance package consisting of SGI Accelerate, SGI MPI, and SGI REACT

• SGI Management Suite: Cluster management software
• Altair® PBS Professional™: Job scheduling and workload management

Software Development

Programming Languages and Debuggers

• C & C++: Intel® C++ Compiler, GNU GCC
• Fortran: Intel® Fortran Compilers (Fortran 95), GNU GCC (Fortran77)

• Debuggers: Intel® Debugger included with Intel® compilers, GNU GDB, Rogue Wave Software TotalView Team, Allinea DDT, Intel® Inspector XE
• Performance Analysis: Intel® VTune Amplifier XE, Intel® Trace Analyzer & Collector

Libraries

• SGI MPI
• Intel® Math Kernel Library
• Intel® Integrated Performance Primitives

• OpenMP included with Intel compilers
• Intel® Parallel Building Blocks
• Intel® MPI Library

 
法律声明 免责声明 加入我们 站点地图 联络我们 中国长城计算机深圳股份有限公司 版权所有 备案编号: 44033110 粤ICP号: 06022800   长城电脑党的群众路线教育实践活动 “小金库”治理专题 Win8