Blue Gene

From TOP500 Supercomputing Sites - Wiki

Jump to: navigation, search

[edit] Description

Blue Gene/L is a computer architecture project designed to produce several next-generation supercomputers, designed to reach operating speeds in the petaflops range, and currently reaching speeds over 280 teraflops (sustained). It is a cooperative project between the United States Department of Energy (which is partially funding the project), industry (IBM in particular), and academia.

Each Compute or IO node is a single ASIC with associated DRAM memory chips. The ASIC integrates two 700 Mhz PowerPC 440 embedded processors, each with a double-pipeline-double-precision Floating Point Unit (FPU), a cache sub-system with built-in DRAM controller and the logic to support multiple communication sub-systems. The dual FPUs give each Blue Gene/L node a theoretical peak performance of 5.6 GFLOPS. It is possible to use only one of the CPUs on a chip for computation and reserve the other one for communication tasks.

Blue Gene/L compute nodes use a minimal operating system supporting a single user program. To allow multiple programs to run concurrently, a Blue Gene/L system can be partitioned into electronically isolated sets of nodes. The number of nodes in a partition must be a positive integer power of 2, and must contain at least 32 nodes.

MTBF time of the largest system installation is about 6.16 days (dominated by memory failure).

[edit] Information Tables

Vendor IBM
System designer IBM
Model name(s) Blue Gene
Time of manufacture 2004
Number of systems sold 158 racks (34 customer sites) as of 2007.08
GENERAL INFORMATION
Class Cluster
Type RISC-based distributed-memory multi-processor
Operating system Linux
Compilers XL Fortran (Fortran 90), XL C, C++
Other specific software ESSL, MASS, MASSV, MPICH2
Affiliated technologies system-on-a-chip
CPU generation used PowerPC 440
CPU model 700MHz CPU (system-on-a-chip)
DRAM type DDR SDRAM 350MHz
FPGA n/a
System cooling Air
Floor space 2500 ft? / ~300tons in full configuration
NETWORK
Node Interconnect Proprietary / 6x175MB/s 3D Torus , 350MB/s Tree
Network switch /
Size / Quantity
-
Network topology 3D torus and tree network
Management network -
BUILDING BLOCKS
Block name --> Node Compute Card Node Board Cabinet System
Structurebasic block 2 Nodes 16 Compute Cards 32 Node Boards up to 64 Cabinets
Number of CPU 2 (in single chip) 4 64 2048 131072
Memory size (max) cache1GB 16GB 512GB 32TB
Other equipment - - 0-2 I/O cards - -
Power consumption 10-15W - - 15-20KW 1.2-1.5MW
Theor.performance 2.8/5.6GF 5.6/11.2GF 89.6/179.2GF 2.9/5.7TF 180/367TF
BENCHMARKS
HPL Benchmark: CPUs /
Rmax / efficiency / Date
- - - - 131072xCPU /280.6TF / 76.45% / 2005
8192xCPU / 18.2TF / 79.4% / 2005
Application performance - - - - -

[edit] External Resourses

Personal tools