• 2013 CitizenShip Report
  • Social Impact on GPU
  • NVIDIA Foundation
Design Advances
Energy-Conscious Design

Whether we are engineering systems to power mobile devices or creating architectures that support high-performance supercomputers, improving energy efficiency is a principle goal in each step of our processes around research and development, and design.

Parallel processing consumes far less power than equivalent computational forms. On a per-instruction basis, GPUs are dramatically more power efficient than CPUs, which have traditionally handled most instructional processing. Our GPUs, for example help the new supercomputer at Italy’s Cineca facility produce 1,500 tons less CO2 output than a comparable CPU-based system.

Our highly efficient products and technologies include:

  • NVIDIA Tesla solutions for high-performance computing: Tesla GPUs deliver the equivalent performance of a multicore CPU at one-tenth its cost and one-twentieth its power consumption. Tesla K20X GPU accelerators power Oakridge National Laboratory’s Titan supercomputer, which is more than five times as efficient as its predecessor and was recently named the world’s fastest computer.
  • NVIDIA CUDA parallel processing architecture: CUDA is a parallel computing platform and computing model that enables compute-intensive calculations to be executed on lower cost, power-efficient GPUs. Learn more about GPU computing.
  • NVIDIA Optimus technology: Optimus maximizes energy conservation and battery life in notebooks by automatically shutting off the GPU when it is not needed.
  • NVIDIA Tegra mobile processors: Tegra 4 quad-core processors utilize 4-PLUS-1 technology, which employs four CPU cores that power up only as needed and a fifth battery-saving core for lower-power tasks. Their PRISM technology reduces a mobile device’s backlight power while simultaneously enhancing the pixel color to deliver the same visual quality with a substantially extended battery life.

With the recent introduction of our Kepler GK110 GPU architecture, our goal was to continue to push the limits of energy efficiency, based on performance per watt. In a recent market test, our Kepler-based GTX 650 cut the system-level power used to run China’s most popular games by 20 to 50 percent.

To reach these levels of efficiency, NVIDIA engineers applied everything learned from the previous Fermi generation of processors to better optimize the Kepler architecture. Our white paper explains specific differences between the previous and current architectures and the innovative engineering that is improving energy efficiency.

NVIDIA Tegra-powered devices use one-tenth the power of a typical notebook and one-hundredth the power of a typical desktop PC.
Mixing NVIDIA GPUs with CPUs makes Oak Ridge Laboratory’s Titan five times as efficient as its predecessor. Read More >