Product Info
Additional Info

GPUs for Defence and Intelligence


The defence and intelligence community heavily rely on accurate and timely information in its strategic and day-to-day operations. Intelligence gathering and assessment are essential parts of these activities that comprises of data coming from a number of diverse sources such as satellites, UAVs, surveillance cameras, and radar. Converting the collected raw data into actionable information requires a significant infrastructure - people, computer HW & SW, power, and facilities - all of which are limited. NVIDIA graphics cards represent a “game changing” technology that dramatically increases productivity while reducing cost, power, and facilities. Using GPUs to augment existing processing systems is an established practice used by high performance computing centres and research institutions around the globe to close the gap between the growing demands of their scientists, engineers and the compute capacity of current IT infrastructure.

In the following charts, we highlight work completed on NPP and CuFFT.

NVIDIA GPU performance NVIDIA GPU performance

Key areas where GPUs are already showing significantly improved performance are:

Image Processing: The role of image processing is expanding for defence and intelligence. The amount of imagery available today for defence and intelligence professionals is unprecedented and collection of new imagery continues every minute. For example, geo-spatial imagery available through satellites already covers the earth’s surface five-fold. There are also over 100 million images of fingerprints stored in the FBI database. GPUs accelerate the image processing workflow including georectification, filtering algorithms, change detection, and 3D reconstruction. Learn more about the impact of GPUs by reading the Digital Globe Case Study on accelerating disaster relief efforts.

Persistent Video Surveillance: Specialists predict that the global video surveillance market will exceed $25 billion by 2016. Additionally each month, the Department of Defence collects more than 10,000 hours of aerial surveillance video in Afghanistan and Iraq. These videos must be processed and analysed in real time. GPUs represent a great tool to archive real time performance for video processing and analytics algorithms.

Signal Processing: The capabilities of modern sensors continues to grow. Making use of the information resources is a growing problem that demands increasing computing capabilities. GPUs are providing a step function in processing speed necessary to keep up with sensor capabilities providing an opportunity for real time integration of sensor data with other data sources to better understand the complex environments our defence teams now operate in. Learn more about the impact of GPUs by reading the OpCoast Case Study on Modeling Radio Jammer Effectiveness.




The video displays the difference in performance between systems with CPUs to a system that includes GPUs. The application calculates Line of Sight to determine visibility for a particular geographic location. The application allows analysts to quickly analyze ground, aerial and radar visibility to determine things like optimal radar placements. The areas with limited visibility are shadowed in green and the areas of high visibility are shadowed in red.


This video shows the analytical process involved in changing low quality videos taken by UAVs into accurate mission critical data that can be used for analysis. Multiple algorithms have to be applied to improve its quality and it clear for analysis. All the algorithms require compute power, especially when computing in real time. On the right, you can see flops required for compute. Once image is clean, one can begin analysis to identify people on the boat and any other moving targets. All is done in real time, which would have been impossible without GPUs.


This video shows real time face recognition made possible by GPUs. The camera captures people walking through the hallway. Images of their faces are automatically matched against an existing database for identification.