Posted on Tuesday, September 21 2010 @ 4:20 CEST by Thomas De Maesschalck
EE Times
reports GE's Intelligent Platforms unit is favoring NVIDIA's CUDA over FPGA-based application development, because it has been demonstrated to be both faster and less expensive. The company has released hardware and software products for NVIDIA's GPUs at an NVIDIA developer conference, it includes an integrated hardware and software development environment, a software environment for CUDA development and debugging as well as a library of optimized math and DSP functions that take advantage of GPGPU computing on NVIDIA's chips.
The software is a set of libraries providing more than 500 digital signal processing and vector math functions optimized for Nvidia's GPUs. It is geared for applications including radar, sonar, image processing and signals intelligence.
“When compared to FPGA-based application development, CUDA has been demonstrated to be both faster and less expensive,” said Rob McKeel, vice president of military and aerospace embedded computing for the GE unit, speaking in a press statement.
