[2] Tera Computer Company's MTA System Stand-Out Performance Subject of 'Best Paper of SC99'.

SEATTLE--(BUSINESS WIRE)--Dec. 7, 1999--

Tera Multithreaded Architecture Praised For

Efficient Parallelization, Programming Ease

And Low Memory Overhead Compared With Competing Systems

Tera Computer Company (NASDAQ NM: TERA), a leading designer of high-performance multithreaded (MTA) computer systems, today announced that the Best Paper of SC99 studied a Tera MTA system and concluded that, compared with competitors, multithreaded systems "offer tremendous potential for quickly and efficiently solving some of the most challenging real-life problems on parallel computers."

The ACM/IEEE SC99 series is the premier forum for high performance computing and networking and drew a near-record 5,100 attendees from around the world this year to its annual meeting.

The "Best Paper" award recognizes excellence in technical contribution to the SC99 conference, and this year's winner was authored by Dr. Leonid Oliker and Dr. Rupak Biswas. Oliker is a post-doctoral scientist in the Scientific Computing Group at the National Energy Research Scientific Computing Center (NERSC), a U.S. Department of Energy facility.

Biswas is an employee of MRJ Technology Solutions and works in the Numerical Aerospace Simulation Division at NASA's Ames Research Center.

Their paper, "Parallelization of a Dynamic Unstructured Application Using Three Leading Paradigms," implemented a mesh adaptation code on the Cray T3E, SGI Origin2000 (NYSE:SGI) and the Tera MTA computers comparing programmability and performance.

The paper reports that, "The success of parallel computing in solving real-life computationally-intensive problems relies on their efficient mapping and execution on large-scale multiprocessor architectures. Many important applications are both unstructured and dynamic in nature, making their efficient parallel implementation a daunting task."

Oliker and Biswas compared several critical factors of parallel code development, including runtime, scalability, programmability, and memory overhead. Their study of the Tera MTA demonstrates that, compared with competitors, multithreaded systems, "offer tremendous potential for quickly and efficiently solving some of the most challenging real-life problems on parallel computers."

The following table summarizes the three different programming paradigms for this study on these state-of-the-art supercomputers. The Tera MTA was the most impressive machine for this dynamic and irregular application. The MTA required a trivial amount of additional code and had little memory overhead. -0-

Program                       Time       Code      Memory
Paradigm       System      In Seconds  Increase   Increase Scalability
-------------- ----------- ---------- ---------- --------- -----------
Serial         R10000       6.40          --         --        --
-------------- ----------- ---------- ---------- --------- -----------
MPI            T3E          3.00      100%       70%       Medium
-------------- ----------- ---------- ---------- --------- -----------
MPI            Origin2000   5.40      100%       70%       Medium
-------------- ----------- ---------- ---------- --------- -----------
CC-NUMA        Origin2000   39.60     10%        5%        None
-------------- ----------- ---------- ---------- --------- -----------
Multithreading MTA          0.35      2%         7%        High
-------------- ----------- ---------- ---------- --------- -----------

-0-

Horst Simon, Director of NERSC Division at Lawrence Berkeley National Laboratory stated, "In addition to operating one of the nation's largest scientific computing centers, NERSC continually researches and evaluates new tools and technologies which may benefit our 2,500 users. Many real-world problems are characterized by irregular data access patterns.

"This makes the research by Oliker and Biswas highly relevant to the needs of researchers and users in many of the Department of Energy's application areas."

Simon added that, "The NERSC has been a partner in an ongoing evaluation of the Tera MTA system installed at San Diego Supercomputer Center with funding from the U.S. Department of Energy."

Burton Smith, Tera's Chief Scientist, said, "This result validates Tera's architectural approach. High performance was achieved on this difficult computational problem with relative ease thanks to the capabilities we designed into the MTA system."

Commenting on the Oliker/Biswas "Best of SC99" paper featuring the Tera MTA performance, Jim Rottsolk, Tera's president and chief executive officer, concluded, "The Tera MTA supercomputer is rapidly building momentum in the marketplace. Being the featured technology, in the most prestigious scientific paper at the world's most impressive gathering of the supercomputing industry, serves as important validation for this technology just as we are beginning to take the Tera MTA into commercialization."

A complete copy of the paper, "Parallelization of a Dynamic Unstructured Application Using Three Leading Paradigms," can be found at http://www.nersc.gov/~oliker/paperlinks.html.

About Tera Computer Company

Tera Computer Company designs, builds and sells high performance general-purpose parallel computer systems. Tera believes its Multithreaded Architecture system represents the next wave in supercomputer technology because of its unique ability to provide high performance, broad applicability and ease of programming in a single system.

For more information about Tera and its MTA systems, contact Tera at 411 First Avenue South, Suite 600, Seattle, WA 98104-2860. Phone: [phone]. Fax: [phone]. E-mail: [e-mail], or http://www.tera.com.

Safe Harbor Statement

This press release contains forward-looking statements, among other things, Tera's plans to build larger MTA systems and the successful running of key applications on the MTA-16. There are certain factors that could cause Tera's execution plans to differ materially from those anticipated by the statements above. Among such factors are risks associated with building larger MTA systems, necessary modifications to software and hardware systems, timely availability of commercially acceptable components from third party suppliers and successful porting of third party applications. For a discussion of such risks, and other risks that could affect Tera's future performance, please see "Risk Factors" in Tera's most recent SEC Form 10-Q.