How relevant is the TOP500 actually and is speed really that important?


17 Jun 2013 Leipzig - Michael Resch from the High Performance Computing Center in Stuttgart in the opening session at the ISC'13 event in Leipzig discussed the pros and cons of the TOP500 and tackled some critical remarks about the relevance of speed in high performance computing. Which meaning does the Olympic motto "Citius, Altius, Fortius" have in the world of supercomputers?

Some people say that speed is not the only thing that matters, Michael Resch started off, but everything we do in HPC has to do with speed.

What is the TOP500 list in fact? It delivers results for the ranking of supercomputers. It is based on one single standard benchmark. It is published twice a year and available since 1993. In this way, Michael Resch defined the famous list.

There are critical voices coming up against some parts in this definition. They claim that there is only one benchmark. Michael Resch thought them to be perfectly right because only one benchmark is not enough.

This fact also has an advantage though. To have more benchmarks would slow down the business, according to Michael Resch.

Others claim that Linpack is the wrong benchmark. This is true, admitted Michael Resch. It only covers a small part of the applications but the other benchmarks do just the same.

Still others claim that the TOP500 distorts the situation in HPC. They say that the TOP500 makes politicians make the wrong decisions but Michael Resch argued that the TOP500 influence is not that big in the market.

According to the critics, the TOP500 also makes scientists make the wrong decisions. In fact, performant games are the most important in the market instead of science.

Experts also claim that the impact of the list on the HPC market is generally too strong. Still, Michael Resch thinks that the Intel market for instance is more interested in mobile phones. The HPC market is not the top in the overall market.

Are there any directors who have made the wrong decision because of the TOP500 list? Michael Resch does not believe it.

Politicians are interested in the ranking but Michael Resch always explains to them why productivity is more important and then politicians forget about ranking and lists.

HPC is very important though and part of the TOP500 is relevant too. The list namely provides a fair comparison of existing results within the limits of the approach. The list also provides a wealth of statistical material. Finally, it serves to highlight the focus of investment by country and field.

So, what can we learn from the TOP500 in the end? We can learn about processors, architectures, and vendors.

Michael Resch explained that between 1993 and 2012, two vendors disappeard from the list. The evolution in the list also shows that the Japanese strategy is not so successful as we think it was.

Michael Resch stated that the TOP500 is one cornerstone of HPC but that the decisions are made by human beings and not by the list. He also insisted that nothing in science is written in stone.

Michael Resch concluded by citing from the novel "Il Gattopardo" by di Lampedusa: "If we want things to stay as they are, we need to change everything."
Leslie Versweyveld