Nvidia reveals more CUDA work

Day three of Nvidia's GTC 2010 event has been and gone, and if you've been following our coverage you'll be staggered to hear that once again the company has concentrated on the growing use of its CUDA GPU offload technology.

The focus of day one was in the use of CUDA in the medical profession, while day two looked at start-up companies making a name for themselves with CUDA-based technologies - so it shouldn't come as a surprise that day three concentrates on yet more CUDA work, this time in the field of high-performance and large-scale computing.

Benoit Meister of Reservoir Labs, showed off a technology his company is working on which promises to make writing code for execution on the GPU significantly easier - a C-to-CUDA mapper. Currently in prototype form, the technology optimises execution and data movement for a wide range of loop code without the programmer having to worry too much about the intricacies of Nvidia's GPGPU coding techniques.

Meister was followed by Nachiket Gokhale, of Weidlinger Associates, who demonstrated his company's NLFLEX technology - which he claims offers an order-of-magnitude increase in simulation speed for shock and blast studies, using supercomputer clusters build around Nvidia GPUs and CUDA programming techniques.

Finally, NICTA's Abbas Bigdeli and Ben Lever showed off a Big Brother-esque large-scale facial recognition system, which harnesses the parallel computing power of GPGPU technology to quickly analyse real-time CCTV footage and alert the operator when a particular face pops into view.

While seeing the growth of CUDA is certainly fascinating, Nvidia's many fans must be wondering one thing: does the company even make cards for gamers any more?