Bill Kramer TOP500 transcript

The TOP500 list has now exceeded its usefulness and actually is causing behavior and decisions that are not helpful to the high-performance computing environment and our users.

One of the top reasons is that the TOP500 list does not encourage good decision making. As a matter of fact, it discourages good decision making, because people are configuring systems that are not as productive to the science and engineering goals that they have in order to rank higher on an arbitrary list.

Another reason is that the metric that is used—high-performance Linpack—no longer represents the wide range of application space that it did 20, 25 years ago. In fact, there are many applications, many uses of computing, that do not at all track to the TOP500 metric, the condensed linear algebra that HPL uses. Instead we have new methods that have evolved that are no longer indicated by ranking on the list, such as sparse matrices, adaptive mesh refinement, graph type problems. And these are disenfranchised—you have no idea about whether a computer will do well with these or not. And they are making up the increasing majority of the science and engineering applications.

Because of this, we at Blue Waters decided to be the first site ever at this scale intentionally not to put our system on the TOP500 list. Blue Waters is all about sustained performance. What we deliver to the science and engineering and other communities to solve their problems in the most timely manner, to solve problems that have never been solved before. The list gives no indication at all about our ability to do that.