How effective are the parallel programming algorithms being used in these systems? 3 million cores are only effective when there are 3 million parallel inputs being processed. Other than real-time image processing / data visualization, at some point, the threads need to be consolidated to develop conclusions. I wonder what the ratio of "idle set-up" time to "productive processor time" is on such a system.
Weather forecasting based on simulation is one of the most demanding computational application for supercomputers (computational fluid dynamics). I am sure we can improve that way better than other forecasters like economists and t-leaves readers!
Greg44, you are so right! Made me laugh when I read your comment on the weather predictions. I used to beat the local weatherman just by talking with a friend in upper NY state. I just asked them what their weather was like (already happened that day and current conditions), then add between 16 and 18 hours and presto our weather here in NH! I would think it would be easier to "predict" the weather if weathermen paid attention to conditions "upstream" rather than computer predictions..
it says 32000 Xeons + 48000 Phis
So, China is going to extend nuclear capabilities via numerical design?
Can't see any other reasonable target for such a massive centralized system (except for prestige, of course).
There was a time when export of technologies, like those Xeon chips, was strictly controlled by the US Commerce Department.
The schoolboy enthusiasm here at EE Times for "fastest" should be overwhelmed by committing actual JOURNALISM and posing the question of how a future nuclear adversary could get a hold of technology to enable the simulation of higher yield nuclear warheads.
Latest ranking at 33
Sunway BlueLight MPP, ShenWei processor SW1600 975.00 MHz, Infiniband QDR
ShenWei SW1600 with DEC Alpha instruction set
One interesting fact is that the first 3 generations of SW processors were 2 years apart, with the latest SW-1600 becoming available in 2010, or 3 years ago.
It is only ridiculous to assume there is no new development of the SW line of processor development.
Replay available now: A handful of emerging network technologies are competing to be the preferred wide-area connection for the Internet of Things. All claim lower costs and power use than cellular but none have wide deployment yet. Listen in as proponents of leading contenders make their case to be the metro or national IoT network of the future. Rick Merritt, EE Times Silicon Valley Bureau Chief, moderators this discussion. Join in and ask his guests questions.