Nvidia’s Hybrid SLIManufacturer: Nvidia
On the Sunday afternoon before CES started, I spent a few hours with Nvidia in order to learn about some of the new products it is planning to release over the next few months – the first of those is a technology known as Hybrid SLI.
During the event, Rob Csongor, Vice President of Corporate Marketing at Nvidia hailed 2008 as the new year of visual computing, because more than ever before there is an increasing demand for GPUs in today’s PCs.
He said that there are many, many different ways to experience visual computing – and whether it’s a discrete GPU, an MCP and GPU, an IGP (or mGPU) or a multi-GPU configuration, developers have to design for the highest common denominator.
The result is a higher cost for the consumer, because not only does the end user have to buy a discrete GPU to experience most 3D games these days, but with that the power requirements also increase – this results in a secondary cost that’s less obvious than the initial financial outlay. What really surprised us during the briefing was that Nvidia’s way to embrace the year of visual computing (as it has coined it) was that, by the end of the first half of this year, it will integrate a GPU into every Nvidia motherboard—on both AMD and Intel platforms.
Click to enlarge
The reason for this is Hybrid SLI, which is an umbrella name for a group of technologies that will become available to end users that have a system with an integrated Nvidia GeForce mGPU and a discrete Nvidia GeForce graphics card that is running on Windows Vista. Over the course of this article, we will outline some of the technology behind Nvidia’s Hybrid SLI technology—including some things that Nvidia didn’t talk about—and to also speculate on where it may go in the future.
The first two technologies that Nvidia is introducing today are HybridPower and GeForce Boost – the company said it is thinking about (and we assume already developing) more technologies, but it is not ready to talk about its future plans for Hybrid SLI. HybridPower is designed to allow you to choose between two different operation modes: maximum performance and low power operation. On the other hand, GeForce Boost combines an mGPU with a discrete GPU in order to deliver higher performance.
Click to enlarge
One thing that we’ve constantly gone on about for the past year or so is how it’s essentially useless for these companies to talk about performance per watt when most of the time the GPUs are sitting using lots of power when they’re idle. That’s anything but performance per watt, isn’t it?
So, the question is – how do you reduce the power to practically zero when you’re not using the GPU? Over the coming pages we’ll talk about how Nvidia plans to address power consumption in a rather aggressive manner.