| Home | Careers | Contact Us | Newsletter | Store |
 |
      |
 |
 |
       |
|
Cray XT4 and XT3 Supercomputers
The Cray XT4 supercomputer, the latest generation massively parallel processor (MPP) system from Cray,
builds on the success of the Cray XT3 system to bring new levels of scalability and sustained performance to high performance computing.
Designed to support the most challenging HPC workloads, the Cray XT4 supercomputer
delivers scalable power for the toughest computing challenges.
Every aspect of Cray XT4 and Cray XT3 systems is engineered to deliver superior performance for massively parallel applications, including:
Like previous MPP systems, the basic building block of the Cray XT4 and Cray XT3 systems is a PE. Each PE is comprised of one AMD processor (single, dual or quad core) coupled with its own memory and dedicated communication resource. This design eliminates the scheduling complexities and asymmetric performance problems associated with clusters of SMPs. It ensures that performance is uniform across distributed memory processes—an absolute requirement for scalable algorithms. Each Cray XT4 and Cray XT3 compute blade includes four compute PEs for high scalability in a small footprint. Service blades include two service PEs and provide direct I/O connectivity. AMD Opteron™ Processors The industry leading AMD Opteron microprocessor offers a number of advantages for superior performance and scalability. The AMD processor's on-chip, highly associative data cache supports aggressive out-of-order execution and can issue up to nine instructions simultaneously. The integrated memory controller eliminates the need for a separate Northbridge memory controller chip, providing an extremely low latency path to local memory—less than 60 nanoseconds. This is a significant performance advantage, particularly for algorithms that require irregular memory access. The 128-bit wide memory controller provides 10.6 to 12.8 GB/s local memory bandwidth per AMD Opteron, or more than one byte per FLOP. This balance brings a performance advantage to algorithms that stress local memory bandwidth. HyperTransport™ technology enables a 6.4 GB/s direct connection between the processor and the Cray XT4 or Cray XT3 interconnect, removing the PCI bottleneck inherent in most interconnects. Scalable Interconnect The Cray XT4 and Cray XT3 systems incorporate a high bandwidth, low latency interconnect, comprised of Cray SeaStar2 chips and high speed links based on HyperTransport and proprietary protocols. The interconnect directly connects all processing elements in a Cray XT4 or Cray XT3 system in a 3D torus topology, eliminating the cost and complexity of external switches. This improves reliability and allows systems to economically scale to tens of thousands of nodes—well beyond the capacity of fat-tree switches. As the backbone of the Cray XT4 and Cray XT3 systems, the interconnect carries all message passing traffic as well as all I/O traffic to the global file system. Scalable Operating System The Cray XT4 and Cray XT3 operating system UNICOS/lc™ is designed to run large complex applications and scale efficiently to 120,000 processor cores. As in previous generation MPP systems from Cray, UNICOS/lc consists of two primary components—a microkernel for compute PEs and a full-featured operating system for the service PEs. The Cray XT4 and Cray XT3 microkernel runs on the compute PEs and provides a computational environment that minimizes system overhead—critical to allowing the systems to scale to thousands of processors. The microkernel interacts with an application process in a very limited way, including managing virtual memory addressing, providing memory protection and performing basic scheduling. The special lightweight design means that there is virtually nothing that stands between a user's scalable application and the bare hardware. This proven microkernel architecture ensures reproducible run-times for MPP jobs, supports fine grain synchronization at scale, and ensures high performance, low latency MPI and SHMEM communication. Scalable Programming Environment Designed around open system standards, the Cray XT4 and Cray XT3 are easy to program. The system's single PE architecture and microkernel-based operating system ensure that system-induced performance issues are eliminated, allowing users to focus exclusively on their applications. The Cray XT4 and Cray XT3 programming environment incudes tools designed to complement and enhance each other, resulting in a rich, easy-to-use programming environment that facilitates the development of scalable applications. The AMD processor's native support for 32-bit and 64-bit applications and full x86-64 compatibility makes the Cray XT4 and Cray XT3 systems compatible with a vast quantity of existing compilers and libraries, including optimized C, C++, and Fortran90 compilers and high performance math libraries such as optimized versions of BLAS, FFTs, LAPACK, ScaLAPACK, and SuperLU. Scalable RAS & Administration The Cray RAS and Management System (CRMS) integrates hardware and software components to provide system monitoring, fault identification, and recovery. An independent system with its own control processors and supervisory network, the CRMS monitors and manages all of the major hardware and software components in the Cray XT4 and Cray XT3 systems. In addition to providing recovery services in the event of a hardware or software failure, CRMS controls power-up, power down, and boot sequences, manages the interconnect, and displays the machine state to the system administrator. Scalable I/O The Cray XT4 and Cray XT3 I/O subsystem scales to meet the bandwidth needs of even the most data intensive applications. The I/O architecture consists of storage arrays connected directly to I/O PEs which reside on the high-speed interconnect. The Lustre file system manages the striping of file operations across these arrays. This highly scalable I/O architecture enables customers to configure the Cray XT3 with desired bandwidth by selecting the appropriate number of arrays and service PEs. It gives users and applications access to a high-performance filesystem with a global namespace. |
 |
Learn More
 Cray XT4 Brochure Cray XT3 Brochure 3D Seismic EarthquakeModels - Pittsburgh Supercomputing Center Calculating SurfaceMass Densities of Galaxy Dark Matter Halos - CSC Climate Modeling -Max Planck Institute for Meteorology Genome-wide Analysisof Human Growth Factors - CSC Sales Inquiry |
|
|
|||||
 |
|||||
|