【24h】

Packaging technology for the parallel supercomputer SX-4

机译:平行超级计算机SX-4的包装技术

获取原文

摘要

The supercomputer SX-4 performs the fastest computational operation in the world and presented in 1994 (FIG. 1). This paper describes the packaging technology, mainly interconnection and cooling technologies of the parallel supercomputer SX-4Firstly, the packaging hierarchy in the SX-4 is described. In the SX-4, performance and system scalability are significantly progressed compared with the former SX-3. As a result, maximum 1 TeraFlops(Floating-point operation per second) performance andsystem scalability from 1 CPU to 32 CPUs were achieved. These achievements largely attribute to the packaging technology of the SX-4. Secondly, the I/O pins connection technology adopted in the SX-4, especially the connection between the card and the backboard is described. This structure requires about 500kg insertion force for card insertion. Precise structure was achieved to insurethe sufficient accuracy limiting the deformation of the parts used in the structure with the help of a FEM(Finite Element Method) analysis. Finally, cooling technology is described. In the SX-4, large amount of heat generated from the LSIs should be cooled off. Therefore, various kinds of techniques were employed. In these efforts, this paper mainly describes a CFD(Computational FluidDynamics) analysis trying to average the distribution of cooling-air velocity so as to normalize the LSI temperature.
机译:超级计算机SX-4在世界上执行最快的计算操作,并于1994年呈现(图1)。本文介绍了包装技术,主要是平行超级计算机SX-4的互连和冷却技术,描述了SX-4中的包装层次。在SX-4中,与前SX-3相比,性能和系统可扩展性显着进行。因此,实现了从1个CPU到32 CPU的最大1 Teraflops(每秒每秒浮点操作)性能和系统可扩展性。这些成就在很大程度上是SX-4的包装技术。其次,描述了在SX-4中采用的I / O引脚连接技术,尤其是卡与背板之间的连接。该结构需要大约500kg插入力的插入力。达到了可靠的精确结构,从FEM(有限元方法)分析中限制了结构中使用的部件的变形。最后,描述了冷却技术。在SX-4中,应冷却来自LSI产生的大量热量。因此,采用各种技术。在这些努力中,本文主要描述了CFD(计算流体动力学)分析,该分析试图平均冷却空气速度分布,以使LSI温度正常化。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号