SGI Origin 2000

SGI Origin 2000
Name	SGI Origin 2000
Developer	Silicon Graphics
Release	1996
Discontinued	2003
Cpu	MIPS R10000
Os	IRIX
Type	High-end server

SGI Origin 2000 The Origin 2000 was a distributed shared-memory server developed by Silicon Graphics that targeted scientific computing, visualization, and large-scale data analysis. It combined scalable hardware interconnects with multiprocessing capability to serve institutions such as national laboratories, research universities, and commercial supercomputing centers. The system bridged needs across academic projects, corporate research, and government initiatives during the late 1990s.

Overview

Introduced by Silicon Graphics alongside contemporaries at technology exhibitions like COMDEX and Supercomputing, the Origin 2000 competed with systems from IBM, Cray, and HP. It was adopted by organizations including Lawrence Livermore National Laboratory, NASA, and Boeing for workloads similar to those run on machines used by Fermilab, Los Alamos National Laboratory, and Argonne National Laboratory. Procurement and deployment involved procurement offices at DARPA, NSF, and the Department of Energy, while academic centers such as Stanford University, MIT, and the University of Illinois used Origin-class machines for projects related to CERN, Caltech, and the University of California system.

Architecture

The Origin 2000 used a distributed shared-memory architecture based on a Scalable Coherent Interface-style interconnect influenced by research from DEC and Carnegie Mellon University. Its NUMA design and directory-based cache coherence drew on concepts explored at MIT, Princeton, and the University of Cambridge, and paralleled efforts at Sun Microsystems and Hewlett-Packard. The interconnect topology and ccNUMA behavior referenced work analogous to that in papers presented at ACM and IEEE conferences, and implementations compared to systems developed by Fujitsu, NEC, and Hitachi.

Hardware Components

Processor modules of the Origin 2000 housed MIPS R10000 CPUs, memory controllers, and cache coherent logic similar to designs evaluated by Intel and Motorola engineers. The chassis accommodated XIO expansion slots compatible with graphics subsystems produced by companies like Evans & Sutherland and 3Dlabs, and I/O adapters from Adaptec and Emulex. Storage and filesystem options interfaced with networked arrays provided by EMC, Seagate, and Quantum, while networking integration supported Ethernet, ATM, and Myrinet technologies used in clusters at Cornell, Princeton, and ETH Zurich.

Operating System and Software

The system ran IRIX, Silicon Graphics’ UNIX variant influenced by System V and BSD traditions from AT&T and the University of California, Berkeley. Software stacks included compilers and libraries from GCC and Intel, scientific packages such as LAPACK, BLAS, and PETSc, and visualization tools associated with OpenGL, Open Inventor, and the Visualization Toolkit used by groups at Los Alamos, Oak Ridge, and JPL. Parallel programming on the Origin 2000 employed MPI and PVM, while profiling and debugging used tools from HP and Sun environments and academic projects at Rice University and the University of Illinois.

Performance and Scalability

Benchmarks for the Origin 2000 were published alongside SPEC, LINPACK, and NAS Parallel Benchmarks run by teams at Sandia, Argonne, and Lawrence Berkeley National Laboratory. Performance scaled with configurations tested against contemporaries such as the Cray T3E, IBM SP series, and NEC SX systems; scalability studies involved workloads from computational fluid dynamics performed by Boeing, molecular dynamics from ETH Zurich, and climate modeling at NOAA. Memory bandwidth and interconnect latency were analyzed in papers from ACM SIGARCH and IEEE HPCA forums comparing results to designs by Microsoft Research and IBM Research.

Models and Configurations

Configurations ranged from small departmental setups to large-scale cabinets deployed at national centers including NERSC and PRACE-affiliated institutions; models offered different boards and enclosures paralleling product lines from Sun Microsystems’ Enterprise servers and HP Superdome systems. Customers in finance such as Goldman Sachs and Morgan Stanley tested midrange configurations for risk modeling, while pharmaceutical companies including Pfizer and GlaxoSmithKline used larger racks for molecular simulations similar to those run at Roche and AstraZeneca.

Legacy and Impact

The Origin 2000 influenced subsequent multiprocessor and NUMA developments at companies like Silicon Graphics later projects, Intel’s NUMA architectures, and research at universities such as UC Berkeley and Stanford. Its integration of high-performance graphics, scalable interconnects, and UNIX-based software stacks informed designs by NVIDIA, AMD, and Cray, and its use in projects at NASA, DOE, and NSF-funded research left traces in scientific workflows still taught at MIT and Caltech. Collections in museums such as the Computer History Museum and exhibits highlighting computing at institutions including the Smithsonian record its role alongside systems from IBM, Cray Research, and DEC. Category:Silicon Graphics