NeuroSolutions CUDA Add-on

NeuroSolutions neural network users can now harness the massive processing power of their NVIDIA graphics cards with the NeuroSolutions CUDA Add-on. Users can see training speed improvements of up to 65 times on neural networks using Levenberg-Marquardt - the most powerful form of back-propagation learning available.

*These NVIDIA cards only perform single-precision computation. The Intel CPU and NVIDIA Tesla card both perform double- precision computations.

Here is a quick comparison of a GPU vs. CPU sample project in NeuroSolutions using three various NVIDIA graphics cards. The graph shows that the simulation run on a traditional (CPU) processor took 3-hours, 27-minutes and 8-seconds to complete while the fastest CUDA-enabled version only took 3-minutes and 16-seconds to complete - that is a 65 time speed increase! Visit the Sample Project page for complete information on the data and test computer used.

So What is CUDA?

CUDA is NVIDIA’s parallel computing architecture that enables dramatic increases in computing performance. With it, applications can harness the power of the proprietary NVIDIA GPU’s (graphics processing units) on their graphics cards. Using CUDA with your NVIDIA graphics card is like having your own number-crunching super computer at your fingertips.

Don’t have an NVIDIA graphics card in your computer? You can replace your existing graphics card with a CUDA-compatible card for under $100. Those who need faster performance may want to consider a higher-end GeForce card for a few hundred more. Those that seek ultimate performance may opt for a professional-grade Tesla card or system, which is specifically designed for intensive number crunching applications.