Intel raised eyebrows: supercomputer will use ARM chips-super computers, Intel-IT information
The world's largest chip maker Intel is struggling because it missed the mobile chip market. Now cell phone chips to Intel's site--supercomputer market.
Super computers used in Government, academia and industry, for a variety of research topics such as nuclear weapons and potential new drugs. Intel chipset driver 90% above, the top 500 most powerful supercomputers, and dominate the server and PC markets. However, smart phones and tablet computers, almost all from the United Kingdom ARM's chips, its main energy efficiency design for a long time.
Fujitsu said this week that it will use processors based on the ARM, creating called Project k Japan supercomputer successor. Fujitsu to Riken Advanced Institute for Computational Science (Riken Institute for advanced computational science) to create the computer, plans to be used in biomedicine, climate and energy research. The computers installed in 2020 is expected to start.
Fujitsu in Germany international Supercomputing Conference announced the news, Intel heard more bad news at the Conference. New list of the most powerful supercomputers in the world, and the last of the new machines are not based on Intel's x86 technology.
China's "Lake of light" systems of manufacturers, the national Supercomputing Center in Wuxi, the use of a Chinese built custom processors of the schema is not specified.
Super computer power can be measured by their number of operations executed per second, the measure known as FLOPS. Lake light of the number of operations per second up to 1 trillion.
Taihu Lake with crazy computing power, but it still wasn't enough. In recent years, the prospects for faster supercomputers have started to slump. Using more powerful chips, usually of Intel to provide high performance computing forecast earnings. But other factors, such as speed of movement of data within a system, has become a limitation. Meanwhile, top super computer for electricity has become a major problem. Build a bigger machine race appears to have hit a wall.
To supercomputers, from mobile devices, chips used by more powerful chip transistors smaller and smaller. This trend is known as Moore's law. However, transistors are "underwater" speed has slowed and chip power out of control. Super computer manufacturers have begun looking for alternative design, can let their growth speed of the machine. One is the ARM.
"This is disruptive for the high performance computing era," Harvard University James Vice President of computing research Cuff said. "Who can design the exact support key algorithm and code, as well as low power machines company, eventually winning in this new game. ”
Starting from 2011, ARM has invested a lot of money, developed for high performance computer chips. It has reached with IBM and a graphics chip maker Nvidia partnership, but the recent partnership, will ensure that the hot software will run on ARM-based processors.
ARM strategy has not been fully tested, because it is not based on the super computer chip design. Jack Dongarra, a Professor of computer science at the University of Tennessee Knoxville pointed out that it is a list of the 500 most powerful supercomputers, one of the authors. But supercomputers new energy-conscious era, it can be successful. "I think the ARM has great potential," he said. "Large machines although it has not yet been demonstrated. But no limit to its use in the design. ”
英特尔侧目：超级计算机将用ARM芯片 - 超级计算机,英特尔 - IT资讯
全球最大的芯片制造商英特尔正在挣扎，因为它错失了移动设备芯片的市场。现在手机芯片进入了英特尔的地盘——超级计算机市场。
超级计算机用于政府、学术界和工业界，用以各种各样的研究课题如核武器和潜在的新药物。英特尔芯片驱动了90%以上的、前500名最强大的超级计算机，同时称霸了服务器和PC市场。但是，智能手机和平板电脑，几乎全部采用来自英国ARM公司的芯片，长期以来其主攻能源效率的设计。
富士通公司本周表示，它将使用基于ARM的处理器，打造名为Project K的日本超级计算机继任者。富士通正在为Riken Advanced Institute for Computational Science（Riken计算科学高级研究所）打造该计算机，计划将其用于生物医学、气候和能源研究。该计算机预计在2020年安装开启。
富士通在德国的国际超级计算大会宣布该消息，英特尔在大会上听到了更多的坏消息。世界上最强大的超级计算机新名单揭晓了，而最后的新机器不是基于英特尔的x86技术。
中国“太湖之光”系统的制造商，位于无锡市的国家超级计算中心，其使用了中国人建造的未指定架构的定制处理器。
超级计算机的功率可以由它们每秒执行操作的数目来测量，该量度被称为FLOPS。太湖之光每秒运算次数达1万亿。
太湖之光具有疯狂的计算能力，但这还不够。最近几年，使超级计算机更快的前景已开始低迷。使用更强大的芯片，通常是英特尔的，以提供高性能计算预测的收益。但其他因素，例如数据可以在系统内移动的速度，已成为一个限制。同时，顶级超级计算机的电费已成为一个大难题。打造更大机器的比赛似乎已碰了壁。
大到超级计算机，小到移动设备，所用的芯片越来越强大，芯片内的晶体管越来越小。该趋势被称为摩尔定律。但是，晶体管的“缩水”速度放缓了，而芯片的功耗开始失控了。超级计算机制造商已经开始寻找替代的设计，可以让他们的机器保持增长的速度。其中之一就是ARM。
“这对高性能计算而言是颠覆性的时代，”哈佛大学计算研究副院长James Cuff说。“那些能设计出准确支持密钥算法和代码、并且功耗小的机器的公司，最终将在这个新的游戏胜出。”
从2011年开始，ARM投入了大量资金，为高性能计算机研发芯片。它已与IBM和图形芯片制造商Nvidia达成合作伙伴关系，而且最近达成的软件合作伙伴关系，将确保热门的研究软件将在基于ARM的处理器上运行。
ARM的策略尚未得到充分测试，因为还没有基于其芯片设计的超级计算机。田纳西州的诺克斯维尔大学的计算机科学教授Jack Dongarra指出，其是500款最强大的超级计算机名单的作者之一。但在超级计算机的新能源意识时代，它可能成功。“我认为ARM有着巨大的潜力，”他说。“虽然它尚未在大型机器演示过。但在设计上没有什么能限制其用途。”