Robert Morton is creating cool vizzes faster than we can post them! Here is his latest view, which visualizes The Top 500 supercomputing sites, as identified and ranked in a biannual list by Select a topic from the table of contents below.

Here is how Robert describes each page:

Supercomputing clusters are used in very different fashions by each country. This dashboard gives an overview of the size and type of each county's supercomputing clusters, and shows how each country has ranked in this list over more than a decade.

China's rapid growth as an economic power is reflected in its compound growth in computing performance. While the United States retains a comfortable lead, China has rapidly risen to second place in overall compute power. In two years it has increased its average rank by 100 positions, and now runs the second most powerful computing cluster.

Moore's Law observes that a microprocessor's transistor count has doubled roughly every two years. Prior to this decade, increasing transistor density correlated with increasing processor frequency. Recently however performance has scaled mostly with parallelism, and we see that processor frequency remains stable while performance per core shows only modest gains.

Increasing the core count of a computing cluster has become the most effective means of increasing computing power. This has driven substantial innovation in networking / interconnect technology, which has kept the per-core computing efficiency relatively stable.

Thanks for the viz and the insight, Robert.

Les ressources suivantes peuvent vous intéresser...

Abonnez-vous à notre blog