By Michael J. Flynn, Oskar Mencer, Veljko Milutinovic, Goran Rakocevic, Per Stenstrom, Roman Trobec, Mateo Valero
Communications of the ACM, Vol. 56 No. 5, Pages 39-42
10.1145/2447976.2447989

Abstract — The race to build ever-faster supercomputers is on, with more contenders than ever before. But, the current goals set for this race may not lead to the fastest computation for particular applications.

Our viewpoint is that the performance metric should become multidimensional measuring more than just FLOPS, e.g., (a) performance per watt, (b) performance per cubic foot, or (c) performance per monetary unit (dollar, yen, yuan, euro, etc.).

As we direct efforts to break the Exascale barrier, we have to ensure that the scale itself is appropriate. A scale is needed: (a) that can offer as much meaning as possible, and (b) that can translate to real, usable, performance, to the highest possible degree. Such a scale should also feature the same two properties, even when applied to unconventional computational approaches.

The full article can be seen on the Communications of the ACM website..