NVIDIA Optimus Discrete Graphics Power-Down Demo

Today we posted a behind the scenes video from an NVIDIA engineering lab here in Santa Clara that captures one of the cool capabilities of Optimus, NVIDIA's new technology that automatically optimizes your notebook for performance and battery life. The video demonstrates Optimus' capability to immediately power on and off the GPU when an application needs it - all while the system is up and running.

The benefit of electrically shutting off components in your notebook, including the GPU, is that it extends battery life. Compared to a standard notebook with a discrete GPU Optimus can extend battery life up to 2x*. The benefit of immediately turning it on again is, of course, that you get all the graphics and processing capabilities of the GPU at a moment's notice. Few people ever get to see this demo because it requires a completely open notebook system - no chassis - just the motherboard, CPU, GPU, Hard drive, and monitor, so it is not exactly portable. This demo is really killer with engineering teams that design notebooks. They practically fall out of their chairs when they see it.

Why? Because with Optimus when the GPU is not needed it is completely powered off automatically and seamlessly WHILE the rest of the system is up and running - the power to the PCI Express bus, the frame buffer memory, the GPU - everything. This is in contrast to switching the GPU to a low-power state or to ‘idle', which would still draw power. You can of course prove the GPU is electrically off by using battery benchmarks and software tests. But the fun way to prove it is to physically PULL THE GPU out of the system while the system is still on and working. NVIDIA

EDITOR'S NOTE: I must admit that this is by far one of the best demonstrations of power-saving technology that the industry has seen. Given that NVIDIA has achieved complete power-down for the GPU with Optimus technology, and not simply accepting the low-power standby state we experience elsewhere, it will be a matter of time before we see this same principal used for other computer hardware components.