• TWEAK
• Gadgets

NVIDIA launches new CUDA libraries

Translate from : NVIDIA lancerer nye CUDA-biblioteker

NVIDIA has launched a series of new CUDA libraries aimed at expanding the capabilities of accelerated computing. According to the NVIDIA Blog, these libraries promise significant improvements in speed and power efficiency across a range of applications.

The new libraries target various applications, including large language models (LLM), data processing, and physical AI. Some of the key highlights include:

Businesses around the world are increasingly adopting NVIDIA's accelerated computing solutions, resulting in remarkable speed increases and energy savings. For example, CPFD's Barracuda Virtual Reactor software, which is used in recycling facilities, runs 400 times faster and 140 times more energy efficient on CUDA GPU-accelerated virtual machines compared to CPU-based workstations.

A popular video conferencing application saw a 66x increase in speed and a 25x improvement in energy efficiency after migrating its live captioning system from CPUs to GPUs in the cloud. Similarly, an e-commerce platform reduced latency and achieved a 33x increase in speed as well as a nearly 12x improvement in energy efficiency by switching to NVIDIA's accelerated cloud computing system.

NVIDIA estimates that if all AI, HPC and data analytics workloads currently running on CPU servers were switched to CUDA GPU-accelerated systems, data centers could save 40 terawatt-hours of energy annually – equivalent to the energy consumption of 5 million US home per year.

Accelerated computing takes advantage of the parallel processing capabilities of CUDA GPUs to perform tasks much faster and more energy efficiently than CPUs. Although adding GPUs increases the maximum power consumption, the total power consumption is significantly lower due to the faster task completion and the subsequent low-power state.

NVIDIA provides a versatile set of libraries optimized for different workloads. The new updates extend the CUDA platform to support a wider range of applications:

• NeMo Curator and Nemotron-4 340B offer advanced capabilities for creating custom datasets and generating high-quality synthetic data.
• cuVS and Polars provide significant performance improvements, enabling large-scale data processing with improved efficiency.
• Warp , Aerial and Sionna introduce new features for physics simulations and research in wireless networks, increasing the capabilities of these platforms.

Essential for accelerating specific workloads, NVIDIA's CUDA libraries offer specialized tools to meet different computational needs. With over 400 libraries, NVIDIA continues to lead the way in delivering powerful and efficient solutions to modern computing challenges.