New FPGA Technology Lets Users Customize Hardware To Accelerate Mission-Critical Applications by Over 100x

MOUNTAIN VIEW, Calif., (Sept. 19, 2005)—Just as supercomputers and clusters changed the face of computing, industry leaders are rallying around an emerging trend aimed at accelerating application performance: reconfigurable computing. To support this industry-wide effort and further extend its high-performance computing (HPC) leadership, Silicon Graphics (NYSE: SGI) today unveiled an advanced hardware solution based on SGI® Reconfigurable Application-Specific Computing (RASC™) technology capable of increasing application performance by hundreds of times over conventional systems.

"SGI's RASC technology offers HPC users a cost-effective way to achieve substantial application performance gains," said Dave Parry, senior vice president and general manager, Server and Platform Group, SGI.

SGI RASC technology enables users to achieve unmatched performance, scalability and bandwidth for data-intensive applications critical to oil and gas exploration, defense and intelligence, bioinformatics, medical imaging, broadcast media, and other data-dependent industries. The new reconfigurable computing technology is available today as an add-in module that seamlessly operates with SGI's Intel® Itanium® 2 processor-based servers and visualization systems.

These data-intensive applications typically run a core set of computing routines, or algorithms, which often consume a majority of total compute time. Traditionally these applications are limited by general-purpose processors that set limits on overall application performance. Increasingly, users are turning to Field-Programmable Gate Arrays (FPGAs) that can be programmed - or reconfigured - by the user for a specific use or application. The FPGA device then serves as a dedicated compute engine for specific routines. Because reprogramming a processor historically required high levels of expertise, FPGA-based acceleration has yet to broadly penetrate HPC markets.

SGI's RASC solution is specifically designed to overcome these challenges by easing the implementation of FPGA technology, making dramatic application performance enhancements available to more users than ever before. SGI's RASC solution set provides a unique combination of capabilities aimed at both enhancing performance and optimizing ease-of-deployment, including:

• An FPGA-aware version of the Gnu Debugger (GDB) built on the current the GDB command set, allowing simultaneous debugging of both the application and the FPGA
• An abstraction layer that enables serial or parallel FPGA scaling
• RASC API and core services library that provides tools to develop reconfigurable computing elements in a multi-user, multi-processing environment
• Collaborative development with third-party HLL tool vendors
• SGI's leadership and technology contributions to OpenFPGA, an industry working group devoted to ensuring that FPGA technologies support powerful HPC and enterprise applications
• Direct connection of the FPGA hardware into the NUMAlink fabric, providing low-latency, high-bandwidth and tight integration of application-specific and general purpose computing elements in a single shared-memory environment
• Virtually limitless scalability of a system's FPGA processing capabilities through the ability to connect multiple RASC expansion modules into a single shared-memory system

Together, these capabilities allow RASC-enhanced instances of SGI® Altix® servers and Silicon Graphics Prism® visualization systems to speed computationally intensive applications by hundreds of times over non-optimized systems.

With long expertise in technical and scientific computing, SGI developed its RASC technology for key applications in several markets. Some examples include the following:

• Oil and gas: time analysis of oil flow and nearly any applications using Fast Fourier Transform algorithms
• Defense/Intelligence: signal processing, edge detection and pattern recognition routines
• Bioinformatics: compare and contrast routines for searching molecule or DNA databases
• Medical imaging: detailed image processing and rendering
• Media: broadcast and post-production transcoding (format conversions), image processing, watermarking, motion estimation, and data conversion

Soaring Performance Gains
SGI customers using systems enhanced with SGI RASC technology already have seen enormous increases in application performance - without the expense of adding more compute nodes or processors. Initial testing has shown a range of applications performance gains ranging from 42x to well over 100x.

To create an ecosystem around RASC that will streamline adoption, SGI has a strategic collaborative arrangement with Nallatech to develop new business opportunities within the HPC market. SGI and Nallatech plan to offer new products and services based on SGI's RASC technology. In addition to Nallatech, SGI has also developed relationships with other industry-leading FPGA technology providers - including Celoxica, Mitrionics, Starbridge Systems, Synplicity and Xilinx - as well as using standard Verilog or VHDL formats - to customize their RASC-equipped SGI systems.

"We're extremely excited by the prospect of a reconfigurable computing solution based on SGI's shared-memory architecture," said Allan Cantle, president and CEO of Nallatech. "Early testing indicates our collaboration with SGI will result in extraordinary performance increases, far surpassing customer expectations."

First milestone for multi-paradigm computing
Riding the current wave of interest among HPC users in FPGA solutions, SGI RASC technology signals a major step forward in the flexibility and capability of industry standard computing systems. Similar to the supercomputing revolution 20 years ago and the emergence of cluster computing systems a decade later, RASC offers the chance to increase application performance by orders of magnitude. Yet unlike these earlier revolutions, SGI RASC technology preserves existing investments in legacy systems by allowing upgraded systems to run a broad range of applications on a single system.

The introduction of SGI RASC technology represents the first major milestone in the company's vision for multi-paradigm computing. A concept pioneered by SGI, multi-paradigm computing enables a single system architecture to meet the needs of a wide array of technical applications. By uniting previously disparate computing architectures with SGI's scalable shared-memory architecture, SGI aims to improve productivity by creating the first supercomputers capable of supporting and combining different computational approaches.

Parry added, "With RASC, SGI combines the performance benefits of using application-specific hardware for core algorithm acceleration with the scalability and ease of use of the NUMAflex global shared-memory architecture to once again push the limits of HPC performance while maintaining a low cost of ownership for our customers. Reconfigurable computing technology is a vital component of our multi-paradigm computing vision and a will be a significant benefit to many of our customers."

SILICON GRAPHICS | The Source of Innovation and Discovery™
SGI, also known as Silicon Graphics, Inc. (NYSE: SGI), is a leader in high-performance computing, visualization and storage. SGI's vision is to provide technology that enables the most significant scientific and creative breakthroughs of the 21st century. Whether it's sharing images to aid in brain surgery, finding oil more efficiently, studying global climate, providing technologies for homeland security and defense or enabling the transition from analog to digital broadcasting, SGI is dedicated to addressing the next class of challenges for scientific, engineering and creative users. With offices worldwide, the company is headquartered in Mountain View, Calif., and can be found on the Web at www.sgi.com.

Silicon Graphics, SGI, Altix, the SGI cube and the SGI logo are registered trademarks, and Silicon Graphics Prism, NUMAlink, RASC, and The Source of Innovation and Discovery are trademarks of Silicon Graphics, Inc., in the United States and/or other countries worldwide. Intel and Itanium are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries. All other trademarks mentioned herein are the property of their respective owners.