SC2000 GORDON BELL AWARDS

THE GORDON BELL PRIZES

Post-conference update: please see the Awards page for a complete list of all awards presented at SC2000.

The Gordon Bell Prizes are awarded each year to recognize outstanding achievement in high-performance computing. The $5000 prize money is donated by Gordon Bell, (research.microsoft.com/~gbell) pioneer in high-performance and parallel computing. A purpose of the award is to track the progress over time of parallel computing in applications. Prizes are awarded in three categories:

  1. Peak Performance: The prize in the peak performance category is given to the entry demonstrating the highest performance achieved in terms of operations per second on a genuine application program. Recent winners have been at or near one teraflop/s.

  2. Price/Performance: The prize in the price/performance category is given to the entry demonstrating the best price-performance ratio as measured in megaflop/s per dollar on a genuine application.

  3. Special: The prize in the special category may be given to an entry whose performance is short of that of the Peak Performance prize, which nevertheless utilizes innovative techniques to produce new levels of performance on a real application. Such techniques may be, for instance, in mathematical algorithms, data structures, or implementations.

Depending on the entries received, in some years no prize may be awarded in a given category. A listing of previous Gordon Bell Awards is available here.

ENTERING THE CONTEST

Submissions for the Gordon Bell Prize competition should be submitted as regular papers to SC2000 and clearly marked for entry in the Gordon Bell prize competition and consideration in one of the three categories above. The Bell Prize committee will select a small number of finalists and notify them at the same time that acceptances for SC2000 papers are announced. Finalists will be able to improve their results during the period allowed for revision of papers accepted for the Technical program. The Prize Committee will select the winners based on the results in the final version of the paper and announce the winners at SC2000.

ADVICE TO SUBMITTERS

Those wishing to enter the competition should familiarize themselves with the high standards set by previous winners. A summary is given below. In addition, one can find more details on past winners by consulting articles published on past competitions. (See Alan H. Karp, Ewing Lusk, and David H. Bailey, "1997 Gordon Bell Prize Winners", IEEE Computer, vol 31, no. 1, January 1998, pp.86-92. Details of the recent winning submissions can be found in the Proceedings of the SC98 and SC99 Conferences.) For all categories, it is important to explain precisely how the performance was measured. For price performance, one must also convince the judges that the price computation is fair and contains no "hidden costs". In the special category, it is the responsibility of the submitters to explain why this computation represents a new level of achievement in its area. For all three categories, it is important that the computation represent a genuine application, not a demo or benchmark. "Embarrassingly parallel" calculations, which require only minimal data communication, have traditionally not been winners. The algorithms used in the calculation must be at least briefly described. An important consideration in judging is whether or not the algorithms used are fundamentally efficient for the given application. Professional practices for citing performance must be observed -- see for example the essay "Twelve Ways to Fool the Masses". Specific questions can be addressed to bellawards@sc2000.org.

Questions: bellawards@sc2000.org


  GORDON BELL AWARDS CHAIR
EWING LUSK, ARGONNE NATIONAL LABORATORY