Cray XT4™ Compute Blade To solve the most challenging scientific and engineering problems, users need scalable computational power. The Cray XT4 blade was designed from the ground up to address a scalable MPI workload. Particularly well suited to problems which require a high-performance interconnect, the Cray XT4 is designed to deliver faster solutions, and meet the demanding requirements of high performance computing (HPC) applications. enables a 6.4 GB/s direct connection between the compute node and the Cray SeaStar2+ interconnect, removing the PCI bottleneck inherent in commodity networks. Memory Each Cray XT4 node can be configured with 1 to 8 GB ECC DDR2 memory. Memory on compute nodes is unbuffered. This feature, Cray XT4 Compute Blade combined with the AMD Opteron’s integrated memory controller provides applications with the lowest possible memory latency. Scalable Interconnect The Cray XT4 blade incorporates a high-bandwidth, low-latency interconnect composed of Cray SeaStar2+ chips and high-speed links based on HyperTransport and proprietary protocols. The interconnect directly connects all compute nodes in a 3D torus topology, eliminating the cost and complexity of external switches. This improves reliability and allows the Cray XT4 system to economically scale to tens of thousands of nodes. As the backbone of the Cray XT4 system, the interconnect carries all message 9.6 GB/sec passing traffic as well as all I/O traffic to the global file system. associated with clusters of large SMPs. It also ensures that performance is 9.6 GB c 9.6 GB/se Cray SeaStar2+ Interconnect /s 9.6 9.6 GB/sec Scalable Compute Nodes Each Cray XT4 blade includes four compute nodes for high scalability in a small footprint. Each compute node is composed of a single AMD Opteron™ processor (dual or quad core), each coupled with its own memory and dedicated Cray SeaStar2+™ communication ASIC. This design eliminates the scheduling complexities and asymmetric performance problems c 9.6 GB/se GB /se c uniform across distributed memory processes – an absolute requirement for scalable algorithms. Each compute node is designed to efficiently run up to four MPI tasks, or alternately can be programmed to run OpenMP within a compute node and MPI between nodes. AMD Opteron The AMD processor’s on-chip and highly associative data cache supports aggressive out-of-order execution. The integrated memory controller eliminates the need for a separate Northbridge memory chip, and provides a high-bandwidth path to local memory – 12.8 GB/sec per single-socket compute node. This design brings a significant performance advantage to algorithms that stress local memory bandwidth. HyperTransport™ technology Upgradeable and Flexible Previous generation Cray XT3 systems can be upgraded to accept the Cray XT4 compute blade. The Linux environment fully supports mixing blade types between system cabinets or even within a single cabinet providing customers with a variety of cost-effective upgrade options. Cray XT4 Specifications CPU 64-bit AMD Opteron series 1000 processors; up to 96 per cabinet Cache 64k L1 instruction cache, 64k L1 data cache, 512KB L2 cache per processor, 2MB unified L3 cache FLOPS 3.5 Teraflops per cabinet Main Memory 1-8 GB Unbuffered ECC DDR2 SDRAM per AMD Opteron Memory Bandwidth 10.6 to 12.8 GB/sec per AMD Opteron Interconnect External I/O Interface 1 Cray SeaStar2+ routing and communications ASIC per Opteron processor 6 switch ports per Cray SeaStar2+ chip, 9.6 GB/s each (57.6 GB/s switching capacity per Cray SeaStar2+ Chip) 3-dimensional torus interconnect Gigabit Ethernet 10 Gigabit Ethernet Fibre Channel (FC) Infiniband (future) Disk Storage Full line of FC-attached disk arrays with support for FC and SATA disk drives File System Lustre file system Cray System Management Workstation (SMW) System Administration Graphical and command line system administration Single-system view for system administration PBS Pro job management system System software rollback capability Reliability Features (Hardware) Cray RAS and Management Subsystem (CRMS) with independent 100 Mb/s management fabric between all system blades and cabinet-level controllers. More than 50 measurement points monitored per Cray XT4 system blade. Full ECC memory protection from memory to system registers Full ECC protection in the Cray SeaStar2+ chip Redundant power supplies Redundant voltage regulator modules (VRMs) Redundant paths to all system RAID Variable-speed blowers with integrated pressure and temperature sensors Simple, micro-kernel-based software design Reliability Features (Software) CRMS system monitors operation of all operating system kernels Operating System Linux environment--Components include SUSE Linux™, Cray Catamount Microkernel, CRMS and SMW software Message Passing Libraries MPI 2.0, SHMEM Compilers Fortran 77, 90, 95; C/C++, UPC Lustre file system object storage target failover; Lustre metadata server failover Software failover for critical system services including system database, system logger, and batch subsystems 15-24.6 kW (15.3-24.85 kVA) per cabinet, depending on configuration. Circuit requirements (Cray XT5 high-density cabinet): 80 AMP at 400/480 VAC (3 phase & ground), 63 AMP at 400 VAC (3 phase, neutral, & ground) Power Cooling Requirement Air cooled, air flow: 3000 cfm (1.41 m3/s); intake: bottom; exhaust: top Dimensions (Cabinet) H 80.50 in. (2045 mm) x W 22.50 in. (572 mm) x D 56.75 in. (1441 mm) Weight (Maximum) 1529 lbs per cabinet (694 kg) Acoustical Noise Level 75 dba at 3.3 ft (1.0 m) Regulatory Compliance UL 60950-1, CAN/CSA - C 22.2 No. 60950-1, CB Scheme Investigation to IEC/EN 60950-1, RoHS, WEEE Safety FCC Class A, DOC Class A, VCCI Class, CISPR 22, EN 50022 Class A, AS/NZS 3548, EN 50082-1, EN 61000-3-2, EN 61000-3-3, Statskontoret 26.2 Category 1 Technical specifications on this page are subject to change without notice. © 2007 Cray Inc. All rights reserved. Specifications subject to change without notice. Cray is a registered trademark, and the Cray logo, Cray XT4, and Cray SeaStar2+ are trademarks of Cray Inc. All other trademarks mentioned herein are the properties of their respective owners.
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project