NVIDIA Updates SDK; Unveils New Technology at GTC

At its annual GPU Technology Conference (GTC), NVIDIA updated its proprietary SDK (software development kit) to expand the Pascal architecture. The goal, the company states, is to make more of its software capabilities available to more developers.

The update includes tools for deep learning, accelerated computing, self-driving cars, design visualization and autonomous machines. They are:

• cuDNN 5: A GPU-accelerated library of primitives for deep neural networks, now includes Pascal GPU support; acceleration of recurrent neural networks, which are used for video and other sequential data; and additional enhancements used in medical, oil & gas and other industries.
• CUDA 8: The latest version of the parallel computing platform, gives developers direct access to powerful new Pascal features such as unified memory and NVLink, NVIDIA states. Also included in this release is a new graph analytics library — nvGRAPH — which can be used for robotic path planning, cyber security and logistics analysis, expanding the application of GPU acceleration in the realm of big data analytics. Critical path analysis has also been added to identify latent bottlenecks in code for CPUs and GPUs.
• An end-to-end HD mapping solution for self-driving cars.
• Virtual reality integration with NVIDIA Iray to let users create VR panoramas.
• NVIDIA GPU Inference Engine (GIE) is a high-performance neural network inference solution for application deployment. Developers can use GIE to generate optimized implementations of trained neural network models.

In addition to these updates to the software development kit, NVIDIA has also released Tesla P100 for next-generation scientific and artificial intelligence and scientific applications.

Additionally, the company introduced the DGX-1 deep learning supercomputer, which is built on Tesla P100 GPUs. According to NVIDIA, it has a 12x increase in neural network training performance from NVIDIA Maxwell, NVLink, 15.3 billion transistors and more than 21 teraflops of peak performance.

For more information, visit NVIDIA.

Sources: Press materials received from the company and additional information gleaned from the company’s website.