SGI® Origin® 3000 systems are the world's most scalable microprocessor-based supercomputers, expandable to over 1,000 processors in a shared-memory environment. The SGI® NUMAflex® shared-memory architecture of SGI Origin 3000 provides large-memory handling and resource management capabilities in a standard, scalable environment for high-productivity supercomputing that is easy to deploy. These systems are designed to meet the high-productivity computing requirements of today's technical professionals. Whether you're predicting global weather patterns, designing blast-resistant buildings, or performing next-generation collaborative surgeries, SGI Origin 3000 ignites your next innovation.
Datasheet (PDF 67K)
End of General Availability for MIPS® IRIX® Products - December 2006
- New 4x-density SGI Origin 3000
- Up to 128 CPUs and 256GB per compute rack
- SGI® NUMAflex® shared-memory architecture
- Easy-to-deploy supercomputing
- SGI® IRGO™ HPC workflow optimization features
Engineered for High-Productivity Supercomputing with SGI NUMAflex Architecture
The latest model in the SGI Origin 3000 family, the SGI Origin 3900, brings deployable, high-productivity supercomputing to your technical application environment with the world's most powerful rack and an operating environment specifically engineered for optimal performance. SGI Origin 3000 supercomputers help drive your breakthroughs with extreme shared-memory scalability and flexibility, all in a deployable computing environment. With our unique SGI NUMAflex memory, you can combine up to 512 CPUs and 1TB of memory in a shared-memory image, enabling breakthroughs that general-purpose business servers just cannot address. The architecture is designed to run complex models frequently, and because the entire memory space is shared, large models fit into memory with no programming model restrictions. This allows system resources to be dynamically reassigned to more complex areas, resulting in faster time to solution--and insight.
As computing requirements continue to increase in today's highly competitive global marketplace, physical space constraints can become an area of growing concern while planning future technology investments. The SGI Origin 3900 server offers deployable supercomputing that eases these constraints, delivering more processors (up to 128), memory (up to 256GB), and application performance per rack than others in the industry, including IBM, HP, or Sun. Following this architectural path, SGI will provide denser, more deployable configurations that can turn broadcast vehicles, petroleum vessels, or aerospace cockpits into computational laboratories.
Exceptionally Low Power Consumption
The SGI Origin 3900 supercomputer capitalizes on the exceptionally low power consumption and heat dissipation of the MIPS® microprocessor line to repackage the existing electronic components and, as a result, quadruple the maximum computational capability per rack. The industry-leading low wattage of MIPS microprocessors has enabled the creation of a new "super-brick" that has up to 16 CPUs and 32GB of memory, plus an SGI® NUMAlink™ router, in a single 4U (seven-inch) module. The new super-brick utilizes the same electronics as previous SGI Origin 3000 systems, enabling easy incorporation into existing systems, as well as simple upgrades.
IRGO: Optimized HPC Workflow Environment
In addition to robust hardware architecture, high-productivity results demand a technical operating environment that is specifically engineered to optimize workflow for greater productivity. IRGO delivers a unique set of IRIX® HPC workflow optimization features designed to give you even more control over your high-productivity supercomputing environment. With IRGO you can produce better code that runs faster in a secure environment, so your breakthrough ideas come sooner.
Customer-Defined Scalable Performance
The SGI NUMAflex architecture is designed for the unique requirements of high-productivity computing. Every HPC application has its own criteria for real performance, so SGI Origin 3900 systems are designed with scalability in mind. With modular components for memory, compute, I/O bandwidth, and visualization tied to a high-bandwidth, low-latency interconnect, you can tailor and evolve your supercomputer to meet your unique performance requirements.