Intelligent 4-Way GPU Advanced Switching Bridge Chip for Simulation and Scene Generation

Award Information
Agency: Department of Defense
Branch: Missile Defense Agency
Contract: W9113M-05-C-0023
Agency Tracking Number: 044-0938
Amount: $99,598.00
Phase: Phase I
Program: SBIR
Awards Year: 2005
Solitcitation Year: 2004
Solitcitation Topic Code: MDA04-142
Solitcitation Number: 2004.4
Small Business Information
CROSSFIELD TECHNOLOGY LLC
4505 Spicewood Springs Road, Suite 360, Austin, TX, 78759
Duns: 135494073
Hubzone Owned: N
Woman Owned: N
Socially and Economically Disadvantaged: N
Principal Investigator
 Gary McMillian
 Technologist & Partner
 (512) 795-0221
 gary.mcmillian@crossfieldtech.com
Business Contact
 David Lamb
Title: Business Administration & Partner
Phone: (407) 491-6929
Email: david.lamb@crossfieldtech.com
Research Institution
N/A
Abstract
The Crossfield simulator will significantly increase the MDA's ability to evaluate a missile defense systems' effectiveness without conducting expensive field tests and will enable faster development of anti-missile systems. The simulator uses clusters of high-performance Central Processor Units (CPUs) and Graphic Processor Units (GPUs) to generate realistic scenes in real-time, and the generated scenes are fed out through electro-optic interface modules interconnected by a low-latency, ultra-high-throughput PCI Express Advanced Switching (AS) fabric. The CPUs support a wide range of fixed-point and floating-point algorithms, and include very high performance vector processors. The GPU internal pipelines are optimized for stream computing, and provide exceptionally high floating-point performance for selected algorithms. GPUs from NVIDIA, ATI and others are becoming increasingly versatile in the types of algorithms they can effectively support, however, while maintaining a significant performance advantage over general-purpose CPUs. Crossfield proposes to develop an intelligent 4-way GPU AS bridge chip to tightly couple clusters of GPUs to an AS fabric. The integrated circuit will include local switching and memory resources configured to support the GPU stream computing paradigm.

* information listed above is at the time of submission.

Agency Micro-sites

US Flag An Official Website of the United States Government