Comments on: IBM announces 20-petaflops supercomputer

By: andmar74

andmar74 — Sat, 19 Nov 2011 10:51:26 +0000

It’s interesting that IBM seems to have a clear path to 100 petaflops. It’s not just about putting in more CPU’s, for example the bandwidth between memory and CPU’s must be large enough so that the CPU’s don’t have to wait.

If a 100 petaflops machine is built soon, there’s just a factor of 10 up to the exaflop machine :-)

By: Giulio Prisco

Giulio Prisco — Sat, 19 Nov 2011 10:04:57 +0000

The IBM announcement has 100 “a scalable peak performance up to 100 petaflops” and “When Blue Gene/Q is fully deployed in 2012 at Lawrence Livermore National Laboratory (LLNL), the system, named “Sequoia”, is expected to achieve 20 petaflops at peak performance.” I think the specific LLNL instance will have a peak of 20 petaflops, and other instances may have even higher performance as Promethean says.

By: scorpio13

scorpio13 — Fri, 18 Nov 2011 19:24:39 +0000

20 pflops or 100 pflops its better than 8 pflops

By: Promethean

Promethean — Fri, 18 Nov 2011 17:09:53 +0000

I think “scalable peak performance” just means they can expand it to 100 PFLOPS in the future by adding more of the same nodes.

By: Fievels

Fievels — Fri, 18 Nov 2011 13:36:29 +0000

Nevermind

By: Fievels

Fievels — Fri, 18 Nov 2011 13:34:09 +0000

… *I think you mean peak performance of 20 petaflops. not 100