Summit vs #1
Man-O-War Cay, 16 Nov 2014 Contact:

This article talks about a new supercomputer slated for deployment in 2017.  I wondered how these new systems might stack up against the field when actually
deployed.  So, here is a plot of the projected "300 peak petaflop/s" (*)
of the Summit system in 2017 (the blue diamond) compared with the's projected #1
system performance at that time (the red line).
It is a little uncanny how close Summit's targetted performance is to the projected "#1" performance.  No doubt not a coincidence.

(*) It looks like plots RMAX.  The article quoted "peak petaflop/s".  From the June 2014 results, IBM's Power-based systems in the top 10 show RMAX/RPEAK of 85%.  (see table to right)
Assuming RPEAK is close to "peak" Flop/s, I have adjusted the 300 "peak petaflop/s" down by 85% in the plot above.  Hopefully, better than ignoring it all together.  Plotted on a log scale, it doesn't make much difference.