Nvidia Announces CUDA 6 Platform, Implements Unified Memory

Technology PC News

nvidia-logo-black

Nvidia has announced its latest wizardry for the PC, the parallel computing platform and programming model CUDA 6. The new platform is said to make parallel programming easier and more efficient with the latest GPUs, accelerating CPU-based applications by up to 8 times. Nvidia has also recently launched the GTX 780Ti video card.

CUDA isn’t only used in fancy 3D animations but is also used in scientific, medical and engineering applications.

Here are the key technical features:

  • Unified Memory – Simplifies programming by enabling applications to access CPU and GPU memory without the need to manually copy data from one to the other, and makes it easier to add support for GPU acceleration in various programming languages.
  • Drop-in Libraries – Automatically accelerates applications’ BLAS and FFTW calculations by up to 8x by replacing existing CPU libraries with GPU-accelerated equivalents.
  • Multi-GPU Scaling – Re-designed BLAS and FFT GPU libraries automatically scale performance across up to 8 GPUs in a single node, totalling over 9 TFLOPS.

Version 6 of CUDA is expected to release in early 2014 and members of the CUDA-GPU Computing Registered Developer Program will be notified when it’s available to download. To join, register here.

 

Related Articles for this post below:

Lost Password