Two Years and No Gain in Supercomputing
-
After more than two years, China's Tianhe-2 super computer remains the fastest in the world with unchanged speed of 33.86Pflops. This is pretty amazing that supercomputing has not moved forward on the top end in such a long time and that China has led the pack by such a wide margin for such a long period of time.
-
Another article on this here: http://www.top500.org/blog/lists/2015/06/press-release/
-
Need some competition for Intel. No matter what you think of AMD at least they keep the pressure on to innovate. We need them (or some other competitor) strong again.
-
Power and Sparc make a good showing too. Intel has plenty of pressure. It's big, and has spot number one, but just in the top ten both IBM and Intel have 40% each, Oracle and AMD have 10% each.
-
@mlnews true, in the SC world they might have some competition - desktop/mobile is driving the innovation right now however. Don't see many IBM nor Oracle powered laptops around heheh.
-
@MattSpeller said:
@mlnews true, in the SC world they might have some competition - desktop/mobile is driving the innovation right now however. Don't see many IBM nor Oracle powered laptops around heheh.
No one needs any that powerful.
-
In that space, it is ARM that is giving them competition.
-
Imagine if Japan's RIKEN supercomputer was to upgrade from its existing SPARC64 VIIIfx to the XIfx!! They are several generations old and still number four on the list. They would instantly gain like 50% or more in clock speed, double their cores or maybe quadruple, increase cache sizes and move to faster architecture and interconnects. That would be a mind-boggling jump just from a CPU upgrade.
-
Fujitsu's own numbers suggest that the upgrade (which I checked, would quadruple the cores) would be expected to result in a 10 fold increase in petaflops.... from 10 to 100!! It could "easily" be three times the performance of the top ranked Tianhe-2.