Fujitsu
Not logged in » Login
X

Please login

Please log in with your Fujitsu Partner Account.

Login


» Forgot password

Register now

If you do not have a Fujitsu Partner Account, please register for a new account.

» Register now

Recent Discussion:

NickBown | 22.12.2018, 14:18
We have come across this issue as well, and don't seem to have found a way around it (the server is ...
NickBown | 20.12.2018, 18:40
Hi everyoneWe've got an RX2540 M1 which won't boot past the Fujitsu splash screen (which shows the i...
MarkM | 15.10.2018, 11:33
Hi there.I do not believe the Fujitsu policy on this subject has changed.So NO you can not order dri...
Nov 28 2015

Top500 List, November 15: Asian Success Continues

The latest edition of the semiannual global Top500 list of supercomputers reveals that within less than six months, the number of ranked Chinese systems has nearly tripled, while once again there were few changes among the Top 10. China's Tianhe-2 remains the world's leading supercomputer for the sixth consecutive time, and two new Cray systems in the US and Germany have joined at #6 and #8.

Based at China's National University of Defense Technology, Tianhe-2 boasts 3.12 million cores and 33.86 petaflops (quadrillions of calculations per second) on the Linpack benchmark. The systems listed in positions 2 through 5 proved to be equally persistent; here we find three U.S.-based installations – Titan (#2), Sequoia (#3), and Mira (#5) – as well as Fujitsu's own K computer (#4). HPC fans and regular readers will easily realize that this ranking order has remained the same for two years on end, and so have the configurations and benchmark results of the Top 5.

For those who like a dash of entertainment with their rankings, positions 6 to 10 offer a few more shifts and even two debutantes: at No. 6 is Trinity, a Cray XC40 systems based on 18.816 third-generation Intel® Xeon® E5-2698 processors with 301,056 cores, an Aries interconnect and Cray's proprietary Linux, which scored 8.1 petaflops sustained maximum performance and could reach a theoretical peak of 11.08 petaflops. Trinity pushed aside Switzerland's Piz Daint, which now takes the seventh spot and remains the highest-ranking European system. Position #8 goes to the second debutante Hazel Hen, another Cray XC40 installation equipped with 15,424 Intel® Xeon® E5-2680 processors with a total of 185,088 cores; otherwise, the setup is the same as Trinity's. Hazel Hen delivered 5.64 petaflops (sustained) and resides at Germany's HLRS, the Höchstleistungsrechenzentrum Stuttgart, a division of the city's university. To be honest, with regard to Hazel Hen the term 'debutante' is a tad misleading – while the system is new among the Top 10, it ranked at #16 and #23 before when it was half its current size. The HLRS success was a bit of a misfortune for Germany's second-highest entry JUQUEEN, which dropped from ninth to eleventh place and came in after Saudi Arabia's Shaheen II and Stampede, which is located at the Texas Advanced Computing Center in Austin.

With Trinity and Hazel Hen added, the majority of the Top 10 systems now run Intel processors; only four systems – Titan, Sequoia, Mira, and the K computer – rely on different suppliers and/or CPU architectures (AMD Opteron, Power BQC, and SPARC64 VIIIfx). Perhaps even more surprising is the fact that only four installations – Tianhe-2, Titan, Piz Daint and Stampede – use GPUs or co-processors as accelerators.

Aside from the Top 10, however, the list has undergone some substantial transformation: China has emerged as a new major player, up to 109 systems from 37 in the previous list. Japan's share has decreased slightly from 40 to 36 systems, thus the overall continental Asian share rises from 107 to 173. Asia's fast success comes at the expense of Europe and the U.S.: since June, the European share dropped by 6.6% to 21.6% or 108 systems (formerly 28.2%/141 installations) – the first decrease after years of a slow but continuous buildup. US-based systems are down to 200, again decidedly less than the 231 from June that already marked an all-time low. These changes seem to be at odds with reports about ongoing stagnation in the field of supercomputing as well as with an assessment from the Top500 editors Erich Strohmaier, Horst Simon (both Lawrence Berkeley National Laboratory), Jack Dongarra (University of Tennessee) and Martin Meuer (Prometeus, Germany), who have noticed a "lag in the overall average performance of all 500 systems" as well as a decline in deployment speed over a period of at least two years. Based on these observations, they conclude that "the market for the very largest systems might currently behave differently from the market of mid-sized and smaller supercomputers." But while this assessment may be accurate, it doesn't reveal much about possible reasons for the 'double slowdown' – and there are plenty that could play a part, from sheer costs to a lack of decent use cases.

For more information, please see the Top500 webpage.

 
SHARE

Comments on this article

No comments yet.

Please Login to leave a comment.