MRI: Acquisition of a Visualization GPU Cluster (VGC)

MRI：获取可视化 GPU 集群 (VGC)

• 批准号: 0521109
• 负责人: Fredric Bernal
• 金额: --
• 依托单位: West Virginia High Technology Consortium Foundation
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2005
• 资助国家: 美国
• 起止时间: 2005-08-01 至 2008-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0521109&HistoricalAwards=false
• 关键词: MRI; Acquisition; Visualization; GPU; Cluster

This project, enabling, enhancing, and supporting high performance parallel computer graphics based research and education, aims at providing researchers and academic institutions access to parallel computer graphics and visualization technology. Visualizaton Graphics Processing Units (GPUs) clusters (VGCs) will be purchased and placed at the Institute for Scientific Research (ISR), Alderson Broaddus College (A-B College), and at West Virginia University (WVU). The VGCs tightly coupled GPUs will allow investigation of algorithms that could not be implemented efficiently in prior systems. As high-speed networking made cluster computing practical, the bandwidth to the GPUs will make cluster visualization practical. Due to slow networking speeds, and even slower display buffer read-backs, prior clusters were forces to operate as a collection of poorly connected PCs. However, the VGC, with its fast GPU-to-GPU connectivity, will be able to operate as a single, integrated entity for performing graphics- and GPU-oriented operations. The acquisition of the VGC will allow the investigation of new visualization and analysis techniques: ultra-high resolution visualizations, image processing of extremely high-resolution video streams, image based reconstruction of complex real-world scenes through multiple video cameras, and the use of GPUs in-the-large to perform scientific calculations and simulations. The VGC provides the high-speed communication required between GPUs. Currently, there are no available systems that scale-up both screen and depth resolution at a reasonable price-performance ratio. Building a VGC can demonstrate that non-specialized commodity, multi-purpose hardware can replace expensive, single-task hardware while offering more functionality and flexibility.Broader Impact: VGCs will be readily available at three institutions to service students, researchers, educators, and small businesses in West Virginia, an EPSCoR state with relatively poor high-technology infrastructure. An organization without access to significant funds or funding sources will be able to learn from this system and software design, and will be able to build its own VGC machine.

该项目支持、加强和支持基于高性能并行计算机图形学的研究和教育，旨在为研究人员和学术机构提供访问并行计算机图形学和可视化技术的途径。可视化图形处理单元(GPU)集群(VGC)将被购买并放置在科学研究院(ISR)、奥尔德森·布罗德杜斯学院(A-B学院)和西弗吉尼亚大学(WVU)。VGC紧密耦合的GPU将允许对在现有系统中无法有效实施的算法进行调查。随着高速网络使集群计算变得实用，到GPU的带宽将使集群可视化变得实用。由于网络速度慢，甚至显示缓冲区回读更慢，以前的集群被迫作为连接不良的PC的集合运行。然而，VGC凭借其快速的GPU到GPU连接，将能够作为单个集成实体运行，以执行面向图形和GPU的操作。收购VGC将允许研究新的可视化和分析技术：超高分辨率可视化，极高分辨率视频流的图像处理，通过多个摄像机基于图像重建复杂的真实世界场景，以及使用大型图形处理器进行科学计算和模拟。VGC提供了GPU之间所需的高速通信。目前，还没有可用的系统能够以合理的性价比同时提高屏幕和深度分辨率。构建VGC可以证明，非专门的商品、多用途硬件可以取代昂贵的单任务硬件，同时提供更多的功能和灵活性。广泛的影响：VGC将在三个机构随时可用，为西弗吉尼亚州的学生、研究人员、教育工作者和小企业提供服务。西弗吉尼亚州是EPSCoR州，高科技基础设施相对较差。一个无法获得大量资金或资金来源的组织将能够从这个系统和软件设计中学习，并将能够建造自己的VGC机器。